diff options
author | Anmol <anmol.karan123@gmail.com> | 2021-08-21 00:39:46 +0200 |
---|---|---|
committer | Jan-Simon Moeller <jsmoeller@linuxfoundation.org> | 2021-08-30 12:14:03 +0000 |
commit | 042806431fafef0ce60ede76c66747e2ebe5d222 (patch) | |
tree | 3c2c070148f86d5eec617f0d3450b5cbd3e9036a /meta-agl-jailhouse/meta-agl-jailhouse.md | |
parent | 46b8ed7b8241ccd211048a91d16dcaf844558919 (diff) |
[PATCH] meta-agl-jailhouse: Add support for virtio over IVSHMEM block
This patch adds the support for the virtio over IVSHMEM block for AGL with Jailhouse.
Current state of the Patchset:
- Add PCI devices from (00:0c.0 - 00:0f.0), for various virtio devices(virtio-ivshmem-{block, console, NET, custom}), primarily for virtio-blk.
- Add Common memory regions for virtio-blk in the Root-Cell and Non-Root-Linux-Cell configurations.
- Add Documentation `meta-agl-jailhouse.md` for usage and Instructions.
- Add `virtio-ivshmem-block` and `virtio-ivshmem-console` recipes for ott Kernel modules.
v2 (Anmol): Update `agl-linux-x86-demo.c` and `qemu-agl.c` for PCI macros and small descriptions for IVSHMEM PCI devices.
V3 (Anmol):
- [RFC] Add packagegroup for the virtio-ivshmem-{block, console} applications.
- Update the `meta-agl-jailhouse.md` for architecture specific information.
v4 (Anmol): Update recipe files according to the new syntax(3.4-honister).
v5 (Anmol):
- Update recipe files for virtio-ivshmem binaries.
- Update `50_local.conf.inc` for `packagegroup-agl-jailhouse`.
v5 (Anmol): Update syntax for `50_local.conf.inc`.
Signed-off-by: Anmol <anmol.karan123@gmail.com>
Change-Id: I1bed4c6f1d3d9b0792a18544ad18edca85a5f4f7
Diffstat (limited to 'meta-agl-jailhouse/meta-agl-jailhouse.md')
-rw-r--r-- | meta-agl-jailhouse/meta-agl-jailhouse.md | 226 |
1 files changed, 226 insertions, 0 deletions
diff --git a/meta-agl-jailhouse/meta-agl-jailhouse.md b/meta-agl-jailhouse/meta-agl-jailhouse.md new file mode 100644 index 00000000..8108efae --- /dev/null +++ b/meta-agl-jailhouse/meta-agl-jailhouse.md @@ -0,0 +1,226 @@ +## Jailhouse support layer + + +Yocto layer that enables the use of the Jailhouse partitioning hypervisor - <https://github.com/siemens/jailhouse>. + +### How to Enable and Use + +### For QEmu x86_64 + +> Note: Right now the below setup is only working for Intel Machines. + +The AGL feature `agl-jailhouse` has to be enabled. That needs to be done when including aglsetup.sh, for example: + +```sh +$ source meta-agl/scripts/aglsetup.sh -m qemux86-64 -b build agl-devel agl-jailhouse +``` + +That will enable this layer and include the `jailhouse` package in the image. + + +Then, in the target system, the cell configurations (*.cell) are placed in `/usr/share/jailhouse/cells/` and the demo inmates (bare-metal applications to run in a non-root cell) are located in `/usr/share/jailhouse/inmates`. + + +After that follow the [AGL-Documentation](https://docs.automotivelinux.org/en/master/#0_Getting_Started/2_Building_AGL_Image/2_Downloading_AGL_Software/) as usual and build the image using `bitbake core-image-minimal`. + + +After successfully building the image we need to QEmulate the Image using `runqemu`: + + +```sh + +$ runqemu qemux86-64 slirp kvm publicvnc serial bootparams="verbose ipv6.disable=1 intel_iommu=off" + +``` + +After successful emulation of the Image, you will see something similar to this: + + +```sh + +[ 0.021231] [Firmware Bug]: TSC_DEADLINE disabled due to Errata; please update microcode to version: 0xb2 (or later) + +[ 0.588075] kvm: already loaded the other module + +[ 0.926525] hdaudio hdaudioC0D0: Unable to bind the codec + +Automotive Grade Linux 11.91.0 qemux86-64 ttyS0 + +qemux86-64 login: root + +``` + +### For Running the Linux Cell + + +In order to spin-up the Root cell and Inmates, you will need Cell Configurations (*.cell) and Inmates, which you will find at `/usr/share/jailhouse/cell` and `/usr/share/jailhouse/inmates`. + + +But for Linux-Cell some preparation is needed and the things which you will require include: + + +- `bzImage` + +- `rootfs.cpio` + + +These files you can find easily in the `build/tmp/deploy/images/qemux86-64/`. + +After obtaining these files you have to `scp` these to the emulated Image. + +Now you should have everything in the target system and you can now spin up the things but before that check, if `jailhouse` is present or not: + +```sh + +qemux86-64:~# lsmod + +Module Size Used by + +jailhouse 36864 0 + +``` +As you can see it’s showing present, In case if it’s not present then run `modprobe jailhouse`, it will load the Jailhouse kernel module. + +After loading the module now we have to enable the Jailhouse-Root-Cell, enable this by the below command: + +```sh + +qemux86-64:~# jailhouse enable /usr/share/jailhouse/cells/qemu-agl.cell + +``` + + And check the console for the logs. + +After loading jailhouse Root-cell, now we have to load the Non-root Linux cell, so for that run the below commands: + + +```sh + +qemux86-64:~# jailhouse cell linux /usr/share/jailhouse/cells/agl-linux-x86-demo.cell bzImage -i rootfs.cpio -w out.file -c "console=ttyS2,115200 earlycon earlyprintk" + +``` + + +> Note: In the above command, when you add the `-w out.file` option then it will spit out some commands to start the non-root cell, if not then the cell will boot as usual. Those spitted out commands would look something like these below, you have to run it one by one: + + + +```sh + +qemux86-64:~# jailhouse cell create /usr/share/jailhouse/cells/agl-linux-x86-demo.cell + +qemux86-64:~# jailhouse cell load linux-x86-demo linux-loader.bin -a 0x0 bzImage -a 0xffc600 rootfs.cpio -a 0x3d89000 out.file -a 0x1000 + +qemux86-64:~# jailhouse cell start linux-x86-demo.cell + +``` + +> Note: As you can see in the spit-out commands there is a `linux-loader.bin` is present, this is a tiny bootloader that is required to boot the Linux-inmate or Linux-non-root cell. It is located in `/usr/libexec/jailhouse/linux-loader.bin` in the Emulated Image. + +After running the above commands you will see that Linux is booting in another console. + +Some helper scripts are present to automate all this, and it can be found [here](https://gerrit.automotivelinux.org/gerrit/gitweb?p=AGL/meta-agl-devel.git;a=tree;f=meta-agl-jailhouse/recipes-extended/jailhouse/files/helper-scripts). + + +### For Testing Virtio over IVSHMEM Block + +You can test the Virtio-block over IVSHMEM by following the steps below: + +**This is for the Root-cell <---> Non-Root-Cell communication.** + +First, you have to boot the Image and enable the `qemu-agl.cell` into the target system, as described in the above sections. + +After that follow the below steps to get it spinning: + +```sh + +qemux86-64:~# jailhouse enable /usr/share/jailhouse/cells/qemu-agl.cell +qemux86-64:~# modprobe uio_ivshmem + +``` + +After this check for the PCI device using `lspci -k`, you will see something like this below: + +```sh + +qemux86-64:~# lspci -k + +. + +. + +. + +00:0c.0 Unassigned class [ff80]: Siemens AG Device 4106 + +Subsystem: Siemens AG Device 4106 + +. + +. + +. + +``` + +After confirming, run the below command, this will create a virtio block backend + +```sh + +qemux86-64:~# echo "110a 4106 110a 4106 ffc002 ffffff" > + +/sys/bus/pci/drivers/uio_ivshmem/new_id + +``` + +And after that, start the backend service like this: + +```sh + +qemux86-64:~# ./virtio-ivshmem-block /dev/uio0 disk.img + +Waiting for peer to be ready... + +``` + +After running the backend-service boot or reboot another cell, and backend will show something like this: + +```sh + +qemux86-64:~ # ./virtio-ivshmem-block /dev/uio0 disk.img + +Waiting for peer to be ready... + +Starting virtio device + +device_status: 0x0 + +device_status: 0x1 + +device_status: 0x3 + +device_features_sel: 1 + +device_features_sel: 0 + +. + +. + +. + +``` + +In the Non-Root-Linux, the disk will show up as /dev/vda and can be accessed normally. + + +>Note: +>- For the `virtio-ivshmem-block` you can find it in target system at `/usr/bin/`, if not then copy it from here `build/tmp/work-shared/qemux86-64/kernel-source/tools/virtio/virtio-ivshmem-block`. +> +>- For the `disk.img`, It needs to be a raw image. But its size matters as that defines the virtual disk size, but you may even leave it empty and only partition or format it from the front-end guest. +> +> e.g: +> +> $ dd if=/dev/zero of=disk.img bs=1M count=1024 +> +> $ mkfs.ext4 disk.img |