frida-server for Android x86 Emulator. possible?

[Semyazz]

How can I prepare frida-server for Android x86 (virtualbox/Genymotion based)?

[oleavr]

(Sorry for the late reply here.)

On a Mac build machine it's compiled like this:

$ make -f Makefile.mac.mk build/frida-android-i386/bin/frida-server

However, there are two issues:

• There's not yet any prebuilt Frida SDK for Android x86. In theory this is just a matter of building one:

$ make -f Makefile.sdk.mk FRIDA_HOST=android-i386

• The Linux injector is missing one tiny bit of logic to handle Android/x86 here. It should be the x86 equivalent of the logic for Android/ARM found here. The existing logic for Linux/x86_64 should be a good guide, as it will be identical to that except using the Android libc API instead of the glibc one.

If you're interested in having a go at porting I'd be happy to help out! (Feel free to drop by #Frida on FreeNode by the way.)

[Semyazz]

thx for the explanation. I'll try to find some time (soon) and port it to Android/x86. I'll let you know how it goes.

[oleavr]

Cool!

[oleavr]

I take it you didn't find time to start on this, so I will give it a try. Just finished updating the build system, and currently building a new SDK for Android/x86. Stay tuned. :)

[oleavr]

There; Android/x86 is now supported in latest git. This will be part of the next Frida release, to be released soon.

[dequis]

:heart:

[oleavr]

:smile:

By the way, it looks like the Dalvik integration doesn't work. I'm afraid I'll have to wrap up this release now, as the weekend is about to end, so let's aim to fix that in the release after this one.

[mehdihz]

Is it now possible to use x86 version of frida? If yes, how can I compile to use it?

[oleavr]

Only the arm version is known to be working correctly, haven't had time to look at the Android x86 port sadly. Feel like taking a stab at it?

[Manouchehri]

@oleavr Not sure if you've seen it yet or not, but Google now ships a x86 image in their emulator. It works miles better than trying to emulate an entire architecture.

Would it be possible to host the x86 binary on build.frida.re?

[oleavr]

@Manouchehri Cool! It used to be on build.frida.re but I removed it after getting bug reports that I didn't have the bandwidth to handle. I can add it back but I could need some help testing it.

[Manouchehri]

I'd be happy to test. Feel free to upload it somewhere else if that's easier.

[oleavr]

Awesome! Two freshly built binaries: http://build.frida.re/frida-snapshot/android/i386/bin/frida-server http://build.frida.re/frida-snapshot/android/x86_64/bin/frida-server

Note that I haven't tested them, so beware of dragons. :)

[Manouchehri]

Nice, thanks! I'll let you know how it goes. Want me to create a new issue tracker or use this one?

[oleavr]

@Manouchehri Feel free to file issues either here in frida-core (for injection issues) or frida-gum for instrumentation issues and Dalvik integration bugs, or frida for anything you're uncertain about.

[devinvisible]

I'm not sure if this is an issue with frida-core or elsewhere but I'm attempting to run the following from my Windows 7 host, hosting a Genymotion - Google Nexus 7 5.1.0 API 22 emuator:

frida-trace -U -i send com.supercell.clashofclans

The output says "Resolving functions..." but it crashes the target app. It also crashes com.android.browser (specifying -i open). My Nexus 7 emulator is running http://build.frida.re/frida-snapshot/android/i386/bin/frida-server (which seems to work just fine). Running frida-ps -U works and shows the target app without troubles.

How can I help you debug this further?

[clviper]

First of all, thnks for the effort on a Android x86 build. I have the same issue that devinvisible . When I run frida-trace -R process and in the fase of Instrumenting functions... the target proccess crashes in the device and I get a connection closed. I am running the server on a Genymotion Google Nexus 4, API 17 and the client on linux.