Install with Qemu Script
This article is a guide to install
Android Generic builds on Qemu. Bliss 14 at the current moment does not work. Gearlock causes resizefs errors that do not occur in the current Project Sakura build. There is a version of Bliss 11.13 that does have a working Gearlock in the BLISS OS telegram group.
Also tested is AOSP 12. with AOSP 12 you may need to boot with hwcomposer.drm to get it to work properly when using virgl graphics. currently default boot does not work, and while using gbm mesa, the android UI crashes when a mouse is drawn. so while a keyboard works. no mouse control even with a real mouse passed through to the VM. this is not an issue when using hwcomposer.drm. I recommend duplicating or editing the hwcompser.drm entry, remove 'DEBUG=2' from it. as it will allow usage as normal.
Qemu while simple in use, can be complicated to understand for beginners. This is because Qemu has a LOT of options for customization, many of which are useful, but more so that are not.
Qemu on windows as of time of writing this will NOT work on bliss 14. but may work on bliss 11. this is due to lack of software rendering on bliss 14. assuming Gearlock issues get fixed. in the mean time, use Project Sakura OS.
If you wish to use this guide on windows 10 or above, It may be possible to use WSL, however to get Qemu KVM support on WSL you will need a custom kernel. Another possible method is to use an unofficial patch from the
qemu-3dfx project to get virgl working on windows. but at this current time. it has not been tested with any android generic project.
Make the image.
When you have Qemu installed be it on Linux or windows, you will have installed a plethora of Qemu related tools. the tool we need for this is
qemu-img, Navigate to the folder you want your image installed in and run the following command.
qemu-img create -f qcow2 Bliss14.qcow2 20G
Breaking this down a little bit running
qemu-img create is used to tell the program we want to create.
-f qcow2 is used to tell it we want to create the image using qcow2 format.
qemu-img supports many formats so we want to specify this one.
Bliss14.qcow2 is the name of the image
20G is the image size.
Make a script to run the VM.
Below is a sample bash script used to run Bliss14 in a Qemu VM
#!/bin/bash qemu-system-x86_64 \ -enable-kvm \ -M q35 \ -m 4096 -smp 4 -cpu host \ -bios /usr/share/ovmf/x64/OVMF.fd \ -drive file=disks/bliss14-k54-gapps.qcow2,if=virtio \ -cdrom images/Bliss14-k54-gapps.iso \ -usb \ -device virtio-tablet \ -device virtio-keyboard \ -device qemu-xhci,id=xhci \ -machine vmport=off \ -device virtio-vga-gl -display sdl,gl=on \ -audiodev pa,id=snd0 -device AC97,audiodev=snd0 \ -net nic,model=virtio-net-pci -net user,hostfwd=tcp::4444-:5555
If you don't want an in-depth explanation, you can skip the next section, Just make sure to replace
-cdrom with the proper disk image, and cdrom image for your use.
While this looks a little complicated, when we break this down we can see that it isn't. we are just telling Qemu what we want our virtual machine to have attached to it. if you don
qemu-system-x86_64 is used to launch the program, there are various other commands that can launch different versions of Qemu, but this is the one we want in the majority of cases.
-enable-kvm tells Qemu to use KVM hypervisor, this is how we get near native CPU performance from our VM
-M q35 This tells Qemu what Machine type to run as, we need to specify this as
i440fx will not work it is simple too old.
-m 4096 -smp 4 tells Qemu How much ram to use, and how many cores to add the the VM
-cpu host This tells Qemu what CPU it should be trying telling the guest it is. Generally it is the preferred option. However in case you start to get instability
-cpu qemu64 may be the option you need as it presents a very bare bones cpu to the VM. see
qemu-cpu-models of the Qemu docs to learn more.
-bios /usr/share/ovmf/x64/OVMF.fd \ is needed to tell Qemu to boot using UEFI, which is necessary as right now there is a bug that prevents android-generic based roms from being installed when in legacy mode. It is important to note that the location of this can vary depending on the host OS.
-drive file=disks/bliss14-k54-gapps.qcow2,if=virtio \ is how we mount the disk. Using this method instead of
-hda lets us use virtio driver instead of emulated driver, giving us greater performance in the VM.
-cdrom images/Bliss-v11.iso \ Just mount the iso for bliss, while we could use virtio driver for this, there is no real need to, because it's only use is installing the OS, this can be removed when the VM is installed
-usb is used to tell Qemu to add a USB controller, no real need to give it any additional arguments
-device virtio-tablet is one of two options for mouse capture the other being
-device virtio-mouse using Tablet will allow you to use Qemu as if it were any other app, using
virtio-mouse on the other hand will capture the mouse, and lock it to the VM. and to free it you will need to click
ctrl+alt to free it. You may use
usb-tablet to use a emulated mouse and tablet instead of the paravirtualized ones. which incurs a very minor preformance hit. but also a measurable latency one.
-device virtio-keyboard is for keyboard, no need to change anything here,
usb-kbd also exists for using emulated keyboard instead.
-device qemu-xhci,id=xhci tells Qemu what USB version to use, no need to change this
-machine vmport=off \ this option turns of vmware I/O emulation. this has caused me some bugs in the past, so I leave it off.
-device virtio-vga-gl this is needed for graphics acceleration, and while you can add a plethora of arguments, none are really beneficial at this time.
-display sdl,gl=on This is used for the display, there are three main options here
display gtk.sdl-on and
-display spice-app,gl=on The difference between these are some features. each one has pros and cons, feel free to experiment. Though do take note that even if you close the VM's window with
-display spice-app,gl=on the VM will still be running, you will need to kill it from the terminal.
-audiodev pa,id=snd0 -device AC97,audiodev=snd0 this is needed to add audio to the VM.
--audiodev pa,id=snd0 is used to create a pulseaudio backend with the id of
snd0. this is important as we then use
-device AC97,audiodev=snd0 to create an emulated AC97 audio device and attach it to the audiodev we created just before.
-net nic,model=virtio-net-pci -net user,hostfwd=tcp::4444-:5555 This really long command is the network command,
-net nic,model=virtio-net-pci is what is used to add the network device to the guest in which case we are using virtio drivers again for best performance.
-net user tells Qemu how to pass the network through and the argument
hostfwd=tcp::4444-:5555 forwards port 4444 and port 5555 together, meaning if we open another terminal and type
adb connect localhost:4444 we can get an adb connection to the VM.
The rest of the installation is the same as real hardware. You should be able to proceed as normal.
Currently we cannot properly set resolution using qemu and virgl. Because of this we need to manually edit the grub file in the VM to do so. There are a plethora of ways to do this. However the easiest is to use qemu-nbd to manually edit the grub entry and use
video=1920x1080 to manually set resoltuon to 1080p. Change this value to a resolution that suits your needs.
Qemu does not currently support vulkan acceleration. While there was a work in progress patch that had support. It has not made it into qemu yet, nor does it seem like it will anytime soon. You may use physical gpu passthrough to get around this issue.
Qemu does not currently support video decode/encode acceleration. There is no news when this might be availible in qemu. You may use physical gpu passthrough to get around this issue.
Qemu is an incredibly powerful tool. this quick guide barely scratches the surface of some of the advanced customizations that are possible. if you are looking for more advanced features to step up your VM. To make it closer to a hardware install. Here are some potential further customizations that are possible with Qemu.
- USB passthrough
- EVDEV usb+mouse passthrough for easy switching
- PCIe passthrough
- egl-headless for remote VM's with graphics acceleration
- emulate USB storage device
- It may also be possible to use multiple virtual monitors, though this has not yet been tested.
This is all possible using virt-manager, or any other libvirt solution given they too support it. no guide has been written up for it yet.