First you need to have giveen's original port installed: http://goo.im/devs/giveen/jellystreak (via the old thread: http://xdaforums.com/showthread.php?t=2130081). The most important thing this does is installing the TWRP "recovery" bootmenu thingy. You can use it when powering on/restarting the dell streak 7 and then keeping power+volup pressed and then choosing "install update from sdcard" or so.
With AOKP there is one install image that wipes /system and an ota update. I have not tested the ota update.
Download for the AOKP 4.2 build for the Dell Streak 7: http://w3studi.informatik.uni-stuttgart.de/~haagch/aokp/
The non-ota update wipes /system. So you have to reinstall gapps every time too, preferably before rebooting (android deletes settings for apps that are not installed I think).
The "official" gapps package uses neon instructions that don't work on tegra2. You'll see the keyboard, tts, etc. crashing all the time. "tonyp" has created a gapps package that uses "old" libraries that work without neon instructions. So you should use this instead of the official gapps:
Download for non-neon gapps: http://goo.im/devs/tonyp/non-neon-gapps
Gesture typing on the keyboard doesn't seem to work for me, but tts works and it doesn't seem to be crashing.
Known issues for me:
[*]sensors don't work: rotation, accelerometer, gps (I think), magnet field(Sensor driver is sensors.p3.so for now, maybe later giveen gets open source drivers to work)- headphone jack doesn't mute/transfer for some headsets like ones with built in microphones
- bluetooth keyboard
- Performance problems. Especially when the ram gets full. You can use a ram manager like https://play.google.com/store/apps/details?id=com.jrummy.apps.memory.manager with the Aggressive or Extreme preset to make that problem go away with the cost of background apps being killed very quickly.
Here is the repository: https://github.com/ChristophHaag/android_device_dell_streak7
And here is how to build it on Archlinux:
AOKP: https://gist.github.com/ChristophHaag/6334554
I'm new to android but maybe some other people know something, so I post whatever I come about. Maybe someone else wants to get started too and finds this helpful.
If you want to engage in bug finding and fixing yourself:
Remote debugging c works like this:
On the android device you do
Code:
gdbserver --remote-debug :5039 --attach 1
For a gui debugger I tried nemiver:
For $ANDROID I use the path where the cyanogenmod was checked out.
Code:
nemiver --remote=<STREAK7-IP>:5039 --gdb-binary=$ANDROID/android/system/prebuilt/linux-x86/toolchain/arm-eabi-4.4.3/bin/arm-eabi-gdb --solib-prefix=$ANDROID/android/system/out/target/product/streak7/symbols/ $ANDROID/android/system/out/target/product/streak7/symbols/init
There's also a statically compiled gdb that you can use over ssh or so: http://dan.drown.org/android/howto/gdb.html
Obsolete first look into the CyanogenMod adb bug:
I think the problem is in line 1068 in init.c
each time I looked when it comes there the cur_action->name was "property:sys.usb.config=none".
Maybe it is connected with the adb issue. When I googled for the
issue I found surprisingly many people having issues with this, but few answers.
But I also found e.g. this: https://gist.github.com/steven676/5...c-remove-obsolete-ro.debuggable-1-trigg.patch
so the problem may be in https://github.com/ChristophHaag/an...lob/master/prebuilts/root/init.streak7.usb.rc
but I didn't have time to really read documentation to that.
I think this file complements $ANDROID/system/core/rootdir/init.usb.rc
I'll either play around with that or I'll add debug output in android.net.LocalSocketImpl.connect(LocalSocketImpl.java:238)
Then I would at least know what it's trying to do and it would get easier.
Many of the results I saw from googling mentioned that it might have to do with netd.
On the streak 7 I get this:
I'm not sure if this is how it should behave...
An observation is that adbd run from a command line seems to start without an issue and listens on a port specified with
but the access over adb connect <STREAK7-IP> does only say "unauthorized". And "start adbd" does nothing. None of the programs seem to have --help or -h, so I have to look closer into whether they can be started directly.
I'll change "[ro.adb.secure]: [1]" in /default.prop to 0 and see whether that does anything.
In the other thread from giveen I said that I don't see the log spam. This was with debugging in the developer settings disabled. When I enable it, the logspam starts. But whether it is enabled or not, init still eats 100% cpu. The trouble with the debugging is that each time it is enabled and I want to disable it, the streak 7 immediately reboots.
Now that I had logcat via ssh running I caught this when the reboot happened:
Looks bad.
For looking at the android code I just use grep and ls with globbing for the c and config files and for the java part I imported it in eclipse via this method: http://source.android.com/source/using-eclipse.html
This is a build I haven't tested yet: http://w3studi.informatik.uni-stuttgart.de/~haagch/cm-10.1-20130820-UNOFFICIAL-streak7.zip
This is giveen's original nvflash that I am not sure I am allowed to put there as giveen has not put any license information in there: http://w3studi.informatik.uni-stuttgart.de/~haagch/JB_Beta2.1.zip But then it's all apache code and there are no notices in reagards to the apache license or changed files anyway. If not, you can just tell me and I'll remove it.
------------------------------------------------
So the call to connectLocal() that throws the exception has as parameters
fd: FileDescriptor[263]
address namespace: RESERVED with address name space id: 1 and address name: adbd
That doesn't help me much yet, but there are frequent calls with FileDescriptor[263] and namespace RESERVED, id 1, name rild (radio service) that don't throw an exception.
So it's a "valid" file descriptor... But I think the problem is still that adbd is not started by init...
The whole UsbDebuggingManager.run method is
where listeToSocket() is ultimately throwing the exception.
I have also read a bit about how adbd is supposed to work. Apparently in android 4.2.2 they introduced rsa encryption. It looks very similar to ssh. You have your authorized public keys on the device in /data/misc/adb/adb_keys (like ~/ssh/authorized_keys) and on your computer you have your public key in ~/.android/adbkey.pub
So I put my ~/.android/adbkey.pub in /data/misc/adb/adb_keys on the streak 7 and started adbd from the command line on the device. And indeed, when I connect with "adb connect <STREAK7-IP>" I get "<STREAK7-IP>:5555 device product:cm_streak7 model:Streak_7 device:streak7" with "adb devices -l" and adb shell works. It's a bit unrelated, but this applies: https://code.google.com/p/android/issues/detail?id=48126
But the actually important part, the "start adbd" still doesn't do anything.
It really must be somethin with /init.streak7.usb.rc. The stuff in /sys/class/android_usb/android0/ seem to be set all wrong...
------------------------------------------------
I'll just keep posting random things I discover that I find strange or interesting and if anyone knows anything about any of those, they can just chime in.
In /init.streak7.usb.rc there is the line
"getprop ro.product.model" says "Streak 7" but /sys/class/android_usb/android0/iProduct apparently doesn't seem to be able to take a string with a space because "cat /sys/class/android_usb/android0/iProduct" returns "Streak". You can write directly to it with "cat "Streak 7" > /sys/class/android_usb/android0/iProduct" but it only saves up to the space. I don't think that's really a problem but strange anyway.
I have googled for another tegra 2 device and looked at its usb init rc: https://raw.github.com/CyanogenMod/android_device_samsung_p4-common/ics/init.p3.usb.rc
Adding a section with on property:sys.usb.config=adb did nothing and it seems I haven't been able to google what should be in /sys/class/android_usb/android0/idProduct for the streak 7.
I think I'll just look into how init on android works and how the triggers work. Then it shouldn't be too hard to figure out what exactly it is trying to do and why it is failing.
------------------------------------------------
I was trying to get my head around why "getprop sys.usb.config" would always return "none" and the system wouldn't respond to "setprop sys.usb.config adb,mtp" in any way. So I got to suspect that /init.streak7.usb.rc was not used at all. Then I compared the imports and found that /init.streak7.rc did use a relative path for /init.streak7.usb.rc while all the other init*.rc were using absolute paths.
So I'm not sure if it this is really the thing that fixed it, but it's the latest thing I tested and now init's 100% cpu and adb are fixed: https://github.com/ChristophHaag/an...mmit/eee0625e11cfafd510c3bada6ae67a133766c0f4
Edit: Wait, it happened again. Maybe not. :/
Hm, no, definitely not it. Can't even reproduce it. It worked after adb sideload and wiping the cache and the dalvik cache.
At least it's clear now that init's 100% cpu usage and adb not working and the dalvik crash when disabling debugging are all the same issue.
Code:
if (!action_queue_empty() || cur_action)
timeout = 0;
Maybe it is connected with the adb issue. When I googled for the
Code:
E/UsbDebuggingManager( 367): Communication error:
E/UsbDebuggingManager( 367): java.io.IOException: No such file or directory
E/UsbDebuggingManager( 367): at android.net.LocalSocketImpl.connectLocal(Native Method)
E/UsbDebuggingManager( 367): at android.net.LocalSocketImpl.connect(LocalSocketImpl.java:238)
E/UsbDebuggingManager( 367): at android.net.LocalSocket.connect(LocalSocket.java:108)
E/UsbDebuggingManager( 367): at com.android.server.usb.UsbDebuggingManager.listenToSocket(UsbDebuggingManager.java:79)
E/UsbDebuggingManager( 367): at com.android.server.usb.UsbDebuggingManager.run(UsbDebuggingManager.java:115)
E/UsbDebuggingManager( 367): at java.lang.Thread.run(Thread.java:856)
But I also found e.g. this: https://gist.github.com/steven676/5...c-remove-obsolete-ro.debuggable-1-trigg.patch
so the problem may be in https://github.com/ChristophHaag/an...lob/master/prebuilts/root/init.streak7.usb.rc
but I didn't have time to really read documentation to that.
I think this file complements $ANDROID/system/core/rootdir/init.usb.rc
I'll either play around with that or I'll add debug output in android.net.LocalSocketImpl.connect(LocalSocketImpl.java:238)
Code:
connectLocal(fd, address.getName(), address.getNamespace().getId());
Many of the results I saw from googling mentioned that it might have to do with netd.
On the streak 7 I get this:
Code:
cat /dev/socket/netd
cat: can't open '/dev/socket/netd': No such device or address
An observation is that adbd run from a command line seems to start without an issue and listens on a port specified with
Code:
setprop service.adb.tcp.port 5555
I'll change "[ro.adb.secure]: [1]" in /default.prop to 0 and see whether that does anything.
In the other thread from giveen I said that I don't see the log spam. This was with debugging in the developer settings disabled. When I enable it, the logspam starts. But whether it is enabled or not, init still eats 100% cpu. The trouble with the debugging is that each time it is enabled and I want to disable it, the streak 7 immediately reboots.
Now that I had logcat via ssh running I caught this when the reboot happened:
Code:
W/dalvikvm( 367): threadid=50: thread exiting with uncaught exception (group=0x40b0e930)
E/AndroidRuntime( 367): *** FATAL EXCEPTION IN SYSTEM PROCESS: UsbDebuggingHandler
E/AndroidRuntime( 367): java.lang.NullPointerException
E/AndroidRuntime( 367): at com.android.server.usb.UsbDebuggingManager.closeSocket(UsbDebuggingManager.java:125)
E/AndroidRuntime( 367): at com.android.server.usb.UsbDebuggingManager.access$200(UsbDebuggingManager.java:46)
E/AndroidRuntime( 367): at com.android.server.usb.UsbDebuggingManager$UsbDebuggingHandler.handleMessage(UsbDebuggingManager.java:177)
E/AndroidRuntime( 367): at android.os.Handler.dispatchMessage(Handler.java:99)
E/AndroidRuntime( 367): at android.os.Looper.loop(Looper.java:137)
E/AndroidRuntime( 367): at android.os.HandlerThread.run(HandlerThread.java:60)
For looking at the android code I just use grep and ls with globbing for the c and config files and for the java part I imported it in eclipse via this method: http://source.android.com/source/using-eclipse.html
This is a build I haven't tested yet: http://w3studi.informatik.uni-stuttgart.de/~haagch/cm-10.1-20130820-UNOFFICIAL-streak7.zip
This is giveen's original nvflash that I am not sure I am allowed to put there as giveen has not put any license information in there: http://w3studi.informatik.uni-stuttgart.de/~haagch/JB_Beta2.1.zip But then it's all apache code and there are no notices in reagards to the apache license or changed files anyway. If not, you can just tell me and I'll remove it.
------------------------------------------------
So the call to connectLocal() that throws the exception has as parameters
fd: FileDescriptor[263]
address namespace: RESERVED with address name space id: 1 and address name: adbd
That doesn't help me much yet, but there are frequent calls with FileDescriptor[263] and namespace RESERVED, id 1, name rild (radio service) that don't throw an exception.
So it's a "valid" file descriptor... But I think the problem is still that adbd is not started by init...
The whole UsbDebuggingManager.run method is
Code:
public void run() {
while (mAdbEnabled) {
try {
listenToSocket();
} catch (Exception e) {
/* Don't loop too fast if adbd dies, before init restarts it */
SystemClock.sleep(1000);
}
}
}
I have also read a bit about how adbd is supposed to work. Apparently in android 4.2.2 they introduced rsa encryption. It looks very similar to ssh. You have your authorized public keys on the device in /data/misc/adb/adb_keys (like ~/ssh/authorized_keys) and on your computer you have your public key in ~/.android/adbkey.pub
So I put my ~/.android/adbkey.pub in /data/misc/adb/adb_keys on the streak 7 and started adbd from the command line on the device. And indeed, when I connect with "adb connect <STREAK7-IP>" I get "<STREAK7-IP>:5555 device product:cm_streak7 model:Streak_7 device:streak7" with "adb devices -l" and adb shell works. It's a bit unrelated, but this applies: https://code.google.com/p/android/issues/detail?id=48126
But the actually important part, the "start adbd" still doesn't do anything.
It really must be somethin with /init.streak7.usb.rc. The stuff in /sys/class/android_usb/android0/ seem to be set all wrong...
------------------------------------------------
I'll just keep posting random things I discover that I find strange or interesting and if anyone knows anything about any of those, they can just chime in.
In /init.streak7.usb.rc there is the line
Code:
write /sys/class/android_usb/android0/iProduct $ro.product.model
I have googled for another tegra 2 device and looked at its usb init rc: https://raw.github.com/CyanogenMod/android_device_samsung_p4-common/ics/init.p3.usb.rc
Adding a section with on property:sys.usb.config=adb did nothing and it seems I haven't been able to google what should be in /sys/class/android_usb/android0/idProduct for the streak 7.
I think I'll just look into how init on android works and how the triggers work. Then it shouldn't be too hard to figure out what exactly it is trying to do and why it is failing.
------------------------------------------------
I was trying to get my head around why "getprop sys.usb.config" would always return "none" and the system wouldn't respond to "setprop sys.usb.config adb,mtp" in any way. So I got to suspect that /init.streak7.usb.rc was not used at all. Then I compared the imports and found that /init.streak7.rc did use a relative path for /init.streak7.usb.rc while all the other init*.rc were using absolute paths.
Edit: Wait, it happened again. Maybe not. :/
Hm, no, definitely not it. Can't even reproduce it. It worked after adb sideload and wiping the cache and the dalvik cache.
At least it's clear now that init's 100% cpu usage and adb not working and the dalvik crash when disabling debugging are all the same issue.
Last edited: