Example: QEMU



As an example, to illustrate potential pain points we will use a Versatile Express CortexA9x4 board, emulated under QEMU (vexpress-a9 target). The board comes with 128MB of CFI NOR flash, provided in form of 2 * 64MB dies. Respective details may slightly differ for NAND flash or SPI NOR flash devices.

It is possible to build a Yocto image and inspect all the details of Mender integration by adding the meta-mender-qemu layer to the build. The layer defines a vexpress-qemu-flash machine and includes all necessary pieces to enable MTD and UBI support.

Testing raw flash support under QEMU requires version >= 2.9 (see the output of qemu-system-arm --version command). Earlier versions contain a bug in CFI flash support that renders flash support on vexpress-a9 unusable.

Add the following to local.conf and run bitbake core-image-minimal:

INHERIT += "mender-full-ubi"


MACHINE = "vexpress-qemu-flash"

A successful build will produce a vexpress-nor image in ${DEPLOYDIR}:

$ ls -shLl tmp/deploy/images/vexpress-qemu-flash/core-image-minimal-vexpress-qemu-flash.vexpress-nor
129M -rw-r--r-- 2 user user 129M 07-18 15:34 tmp/deploy/images/vexpress-qemu-flash/core-image-minimal-vexpress-qemu-flash.vexpress-nor

A vexpress-nor image is a tar file that contains an image for each of the nor 'drives' emulated by QEMU. It is specific to QEMU and will not generally be used for other devices, which will usually use either ubimg or mtdimg.

The image can be run by calling a mender-qemu helper script provided in meta-mender-qemu layer:

QEMU_SYSTEM_ARM=$HOME/qemu-install/bin/qemu-system-arm \
DISK_IMG=tmp/deploy/images/vexpress-qemu-flash/core-image-minimal-vexpress-qemu-flash.vexpress-nor \
MACHINE=vexpress-qemu-flash \

The QEMU_SYSTEM_ARM environment variable can optionally provide a path to qemu-system-arm binary used by the script. When empty or not defined, the script will call qemu-system-arm available in your $PATH.

We welcome contributions to improve this documentation. To submit a change, use the Edit link at the top of the page or email us at .