Re: [Qemu-devel] [PATCH RFC v4 3/4] docker: Support "QEMU_CHROOT" in dockerfiles

[Paolo Bonzini <pbonzini@redhat.com>]

On 11/07/2016 12:08, Alex Bennée wrote:
> 
> Paolo Bonzini <pbonzini@redhat.com> writes:
> 
>> On 11/07/2016 05:20, Fam Zheng wrote:
>>> This allows a docker file to say "ENV QEMU_CHROOT /path/to/new/root" to
>>> indicate that the test execution should be done in a chroot in the
>>> container.
>>>
>>> Bind mount dev,sys,proc into QEMU_CHROOT to make them avaiable for
>>> testing scripts.
>>>
>>> The SYS_ADMIN is a required capability for mount, add it to the
>>> docker run command line.
>>>
>>> Signed-off-by: Fam Zheng <famz@redhat.com>
>>> ---
>>>  tests/docker/Makefile.include |  1 +
>>>  tests/docker/run              | 12 ++++++++++++
>>>  2 files changed, 13 insertions(+)
>>>
>>> diff --git a/tests/docker/Makefile.include b/tests/docker/Makefile.include
>>> index c5546ee..e9821ba 100644
>>> --- a/tests/docker/Makefile.include
>>> +++ b/tests/docker/Makefile.include
>>> @@ -107,6 +107,7 @@ docker-run-%: docker-qemu-src
>>>  		$(call quiet-command,\
>>>  			$(SRC_PATH)/tests/docker/docker.py run $(if $V,,--rm) \
>>>  				-t \
>>> +				--cap-add SYS_ADMIN \
>>>  				$(if $(DEBUG),-i,--net=none) \
>>>  				-e TARGET_LIST=$(TARGET_LIST) \
>>>  				-e EXTRA_CONFIGURE_OPTS=$(EXTRA_CONFIGURE_OPTS) \
>>> diff --git a/tests/docker/run b/tests/docker/run
>>> index 38ce789..4e80cc3 100755
>>> --- a/tests/docker/run
>>> +++ b/tests/docker/run
>>> @@ -19,6 +19,18 @@ fi
>>>
>>>  BASE="$(dirname $(realpath $0))"
>>>
>>> +# cp files into the chroot and execute there
>>> +if test -n "$QEMU_CHROOT"; then
>>> +    mkdir -p $QEMU_CHROOT/$BASE
>>> +    cp $BASE/* $QEMU_CHROOT/$BASE
>>> +    QEMU_CHROOT_SAVE="$QEMU_CHROOT"
>>> +    for bp in dev sys proc; do
>>> +        mount --bind /$bp $QEMU_CHROOT/$bp
>>
>> Can you ask docker to do these bind mounts instead?
> 
> AFAICT docker's various mount directives are all focused on tasks like mounting data
> volumes from the host into the container.
> 
> It's a bit of a shame having to do this as the original approach was to
> use docker to avoid having fancy bind mounts on my hosts system. We are
> now getting to inception levels of nesting here. But the benefit is not
> requiring the host having the pre-requisites to bootstrap the system.
> 
> That said looking at the debootstrap requirements I've seen instructions
> that start with download the deb, ar extract and then run the script by
> hand so maybe this is over complicating things?
> 
> Is it possible to bootstrap a Fedora rootfs with a similar script?

In theory "yum" is all that you need to bootstrap a Fedora rootfs.

Paolo