quic / gunyah-hypervisor Goto Github PK

Thank you!

Hi,

would it be possible to run Gunyah on a Qualcomm RB5? Do you intend to release a configuration to support this?

Thank you,
Gergely

The CPU is Snapdragon 8 Gen 2 and the Tab S9 Ultra has 16 GB of memory. I am trying to find ways to use this device to run VMs. I would like to use Gunyah but I don't know how to install it.

I have built the environment using provided Dockerfile on a Ubuntu 16.04 machine without errors. The image is on dockerhub: ml2558/gunyah:gunyah. However, when I follow the rest process building the hypervisor, resource manger and c runtime and pack them into single hypvm.elf binary and try to run them using qemu, it does not have any output. Is it normal?

Edit: here is what it looks like:

(gunyah-venv) root@a12fdfadd046:/home/gunyah-hypervisor# qemu-system-aarch64 -machine virt,virtualization=on,gic-version=3,highmem=off \

-cpu max -m size=2G -smp cpus=8 -nographic
-kernel /home/hypvm.elf
-device loader,file=$OUTPUT_DIR/Image,addr=$LINUX_BASE
-device loader,file=$OUTPUT_DIR/virt.dtb,addr=$DT_BASE
-device loader,file=$OUTPUT_DIR/initrd.img,addr=$INITRD_BASE

after entering the above command, the terminal just hangs (no errors, crashes, etc).

1. While debugging, I found that pvm KASLR is enabled:
   [ 0.000000] KASLR enabled

2. PVM HLOS mem info
   [RM]HLOS Mem Base : 40000000
   [RM]HLOS Mem Size : 40000000
   [RM]HLOS IPA base : 40000000
   [RM]HLOS DT Ofst : f00000
   [RM]RAM FS offset : 800000

3. start_kernel addr
   root@qemuarm64:# cat /proc/kallsyms |grep start_kernel
   ffffde1763e27d0c T secondary_start_kernel
   ffffde1764fb03c0 T start_kernel
   root@qemuarm64:# cat /proc/kallsyms |more
   ffffde1763e00000 T _text

I tried to set a breakpoint, but it didn't work:
add-symbol-file ./../core-utils/build/vmlinux 0xFFFFDE1763E00000
b start_kernel

Hi, I have a question similar to #15

I succeeded in Gunyah's non-docker setup, but only the primary VM seems to be running.
I found the documentation (https://github.com/quic/gunyah-hypervisor/blob/develop/CHANGELOG.md) suggesting a Linux patch and CrosVM VMM for SVM loading.

Is there any documentation I can follow to load the secondary VM on Gunyah?
I also wonder if Gunyah can load a secondary VM without CrosVM or AVF.

Any advice would be appreciated. Thank you.

1. When launching the Secondary Virtual Machine (SVM) using Crosvm, I encounter the message "No soundcards found" despite having recompiled the Linux kernel (Image) with the CONFIG_SND_VIRTIO=y flag enabled. Does Crosvm lack support for VirtIO sound, or are additional backend configurations required in the Primary Virtual Machine (PVM) to enable VirtIO sound?

2. Instead of Crosvm, can I use Libvirt/Virt-install in the PVM to create and launch the SVM? If yes, would I need to install Libvirt/Virt-install on the PVM? Note that the PVM does not support installing packages via apt.

There are no shared ".githooks" files:
.githooks/pre-commit
.githooks/commit-msg
mentioned in Contribution page
(same with C Runtime and Resource Manager)

I'm teaching a class at UCSD Ext and thinking of using Gunyah for students to play around.
Some are Windows users and only have WSL - Windows Subsystem for Linux.

We could not get "docker build -f gunyah/Dockerfile -t user:gunyah ." to complete because the gcc compiler or ninja ran out of memory on WSL.

Is it possible to host a prebuilt docker instance on hub.docker.com? I found a lot of Android and KVM dockers, but no Gunyah docker there.

Hello, is it possible to support to run RTOS with gunyah-hypervisor or any plan?

Following the Guide, I can successfully run gunyah and the VM, but I want do some more study about it,
so, I want to add A Test App( A image) into the VM, then I can run some test.
But I failed.
I tried the below two ways:

1. add the image as the qemu-system-aarch64 with "--drive " option like below:

qemu-system-aarch64 \
-machine virt,virtualization=on,gic-version=3,highmem=off \
-cpu max -m size=2G -smp cpus=8 -nographic \
-kernel ../gunyah/gunyah-hypervisor/hypvm.elf \
-device loader,file=/usr/local/src/out/Image,addr=0x41080000 \
-device loader,file=/usr/local/src/out/virt.dtb,addr=0x44200000 \
-device loader,file=/usr/local/src/out/initrd.img,addr=0x44400000 \
**-drive  format=raw,index=0,file=../gunyah/myAPP.squashfs**

The command runs successfully without error, but after login to VM OS, there nothing can be seen in any folders in the VM OS

2. Try add my APP into the initram disk image, and got error below:
   [ 36.559017] Initramfs unpacking failed: read error
   .......
   .....
   [ 55.419495] Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(1,0)
   [ 55.435487] CPU: 5 PID: 1 Comm: swapper/0 Not tainted 5.14.0-rc1-00199-g45312bd762d3-dirty
   [ 55.441980] Hardware name: linux,dummy-virt (DT)
   [ 55.445334] Call trace:
   [ 55.448231] dump_backtrace+0x0/0x1d4
   [ 55.461203] show_stack+0x1c/0x28
   [ 55.467446] dump_stack_lvl+0x68/0x80
   [ 55.480107] dump_stack+0x18/0x3c
   [ 55.481206] panic+0x17c/0x36c
   [ 55.482637] mount_block_root+0x1d4/0x200
   [ 55.484015] mount_root+0x60/0x6c
   [ 55.488445] prepare_namespace+0x11c/0x198
   [ 55.489134] kernel_init_freeable+0x154/0x184
   [ 55.489837] kernel_init+0x24/0x19c
   [ 55.490759] ret_from_fork+0x10/0x18
   [ 55.494631] SMP: stopping secondary CPUs
   [ 55.514051] Kernel Offset: 0x4396b0c80000 from 0xffff800010000000
   [ 55.517683] PHYS_OFFSET: 0x40000000
   [ 55.524732] CPU features: 0x000002d1,6d300c47
   [ 55.533291] Memory Limit: none
   [ 55.552053] ---[ end Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(1,0) ]---

Any advice would be greatly appreciated.

Following the steps from https://github.com/quic/gunyah-hypervisor/blob/develop/docs/test.md,

I can boot the Gunyah Hypervisor with the Linux VM on the QEMU simulator.

But it seems there is just one rootvm.

Can we run two VMs on the Gunyah Hypervisor with the Linux VM on the QEMU simulator now ?

Any suggestions would be much appreciated. Thanks

As far as I know, Gunyah is currently working on Qemu support. I wanted to ask if you could provide compiled Qemu with Gunyah support. I have tried several times to compile the static version of Qemu using the instructions provided in the document, but without success because the compilation environment lacks some of the static version libraries.

./configure --target-list=aarch64-softmmu --enable-debug --enable-gunyah --static

https://patchew.org/QEMU/[email protected]/

Hello gunyah team,

Where is the definition of the trigger_vectors_* function? For example trigger_vectors_trap_data_abort_el2_event.

Thanks

There is no documentation for project architecture. Need architecture-specific documentation to start using the hypervisor

Hi,

I get the following error while building the resource manager. Any pointers would be very helpful.

cheers,
Achin

[50/50] ld build/resource-manager
FAILED: build/resource-manager
/home/achgup01/work/tools/llvm/llvm-musl-install/bin/clang -target aarch64-linux-gnu -fuse-ld=lld -mcpu=cortex-a75 -static-pie -Wl,--gc-sections -static -fuse-ld=lld --rtlib=compiler-rt -Wl,-z,max-page-size=4096
-Wl,-z,separate-loadable-segments -lfdt -Wl,-Bstatic -L /home/achgup01/work/tools/sysroot/lib --sysroot=/home/achgup01/work/tools/llvm/llvm-musl-install//aarch64-linux-gnu/libc/ build/arch/armv8/obj/src/smc.c
.o build/platform/qemu/obj/src/uart_qemu.c.o build/src/obj/dt/dt_overlay.c.o build/src/obj/event/event-isr.c.o build/src/obj/exit/exit.c.o build/src/obj/guest_interface.c.o build/src/obj/hyp/memextent.c.o build/
src/obj/irq_manager/irq_manager.c.o build/src/obj/memparcel/memparcel.c.o build/src/obj/preempt/preempt.c.o build/src/obj/resource-manager.c.o build/src/obj/rpc/rm-rpc-fifo.c.o build/src/obj/rpc/rm-rpc-msgqueue.
c.o build/src/obj/rpc/rm-rpc.c.o build/src/obj/uart/uart.c.o build/src/obj/utils/address_range_allocator.c.o build/src/obj/utils/circular_buf.c.o build/src/obj/utils/dict.c.o build/src/obj/utils/vector.c.o build
/src/obj/vm_config/dtb_parser.c.o build/src/obj/vm_config/vm_config.c.o build/src/obj/vm_config/vm_config_parser.c.o build/src/obj/vm_config/vm_get_resources.c.o build/src/obj/vm_console/vm_console_simple.c.o bu
ild/src/obj/vm_creation/dto_construct.c.o build/src/obj/vm_creation/hlos_vm.c.o build/src/obj/vm_creation/second_vm.c.o build/src/obj/vm_creation/vm_creation.c.o build/src/obj/vm_mgnt/vm_mgnt.c.o -o build/resour
ce-manager
ld.lld: error: cannot preempt symbol: __init_array_start

defined in
referenced by __libc_start_main.c:13 (../src/env/__libc_start_main.c:13)
__libc_start_main.o:(.rodata+0x18) in archive /home/achgup01/work/tools/llvm/llvm-musl-install//aarch64-linux-gnu/libc//lib/libc.a

ld.lld: error: cannot preempt symbol: __init_array_end

defined in
referenced by __libc_start_main.c:13 (../src/env/__libc_start_main.c:13)
__libc_start_main.o:(.rodata+0x20) in archive /home/achgup01/work/tools/llvm/llvm-musl-install//aarch64-linux-gnu/libc//lib/libc.a

ld.lld: error: cannot preempt symbol: _DYNAMIC

defined in
referenced by __init_tls.c:80 (../src/env/__init_tls.c:80)
__init_tls.o:(.rodata+0x0) in archive /home/achgup01/work/tools/llvm/llvm-musl-install//aarch64-linux-gnu/libc//lib/libc.a

ld.lld: error: cannot preempt symbol: __fini_array_end

defined in
referenced by exit.c:15 (../src/exit/exit.c:15)
exit.o:(.rodata+0x0) in archive /home/achgup01/work/tools/llvm/llvm-musl-install//aarch64-linux-gnu/libc//lib/libc.a

ld.lld: error: cannot preempt symbol: __fini_array_start

defined in
referenced by exit.c:15 (../src/exit/exit.c:15)
exit.o:(.rodata+0x8) in archive /home/achgup01/work/tools/llvm/llvm-musl-install//aarch64-linux-gnu/libc//lib/libc.a
clang-10: error: linker command failed with exit code 1 (use -v to see invocation)
ninja: build stopped: subcommand failed.

Is it possible to work in Android?
My device chipset Qualcomm sm8350(888)
In my device have hyp partition
My kernel version 5.4
I found some data in kernel device tree

https://github.com/pjgowtham/android_kernel_oneplus_sm8350/blob/ffa3efd23c005672bff907f71a619dfd9cf82aec/arch/arm64/boot/dts/vendor/qcom/lahaina.dtsi#L402

error_t
vic_bind_hwirq_forward_private(vic_t *vic, hwirq_t *hwirq, virq_t virq)
{
error_t err = OK;

assert(hwirq->action == HWIRQ_ACTION_VIC_BASE_FORWARD_PRIVATE);

LOG(ERROR, WARN, "vic_bind_hwirq_forward_private virq {:x}", (unsigned long)virq);

// allocate for private forward
struct partition *partition = vic->header.partition;
assert(partition != NULL);

size_t size = sizeof(hwirq->vic_base_forward_private[0]) * GIC_PPI_NUM;  //here, GIC_PPI_NUM shoud be cpu count.
    //let's see how to use it later
   //		vic_forward_private_t *forward_private =
   //   	      &hwirq->vic_base_forward_private[pcpu];  
  // Am I right?


void_ptr_result_t alloc_r = partition_alloc(
	partition, size, alignof(hwirq->vic_base_forward_private[0]));
if (alloc_r.e != OK) {
	err = ERROR_NOMEM;
	goto out;
}

Hi,

I encounter a system hang while testing the Gunyah Hypervisor with the Linux VM on the QEMU simulator.
(Follow the https://github.com/quic/gunyah-hypervisor/blob/develop/docs/test.md)

Any suggestions would be much appreciated. Thanks

[RM] Starting Resource Manager, version: 727cc85 (Tue Dec 21 03:28:12 2021 UTC)
[RM] memparcel_construct sglist:
[RM] [0]: me:fabfdee110ae6499 phys:40000000 (20000000)
[RM] memparcel_construct ret=0
[RM] create_dtbo: hyp -1 resmem -1
[RM] create_resmem_nodes: memparcel labelled 0 is the base memory at 0x40000000
[RM] init completed
[RM] UART is disabled
[HYP] sysreg_read: unhandled TID3 trap, ISS: 0x34004d. RAZ
[HYP] sysreg_read: unhandled TID3 trap, ISS: 0x380089. RAZ
[HYP] sysreg_read: unhandled TID3 trap, ISS: 0x3a0089. RAZ
[HYP] sysreg_read: unhandled TID3 trap, ISS: 0x3a0007. RAZ
[HYP] sysreg_read: unhandled TID3 trap, ISS: 0x3c0007. RAZ
[HYP] sysreg_read: unhandled TID3 trap, ISS: 0x380007. RAZ
[HYP] Unhandled trap from VM 3, ESR_EL2 = 0x62380405, ELR_EL2 = 0xffffa9bd8b497174