The runc-docker has all the code in it to properly run a stop hook if
you use it in the foreground. It doesn't work in the back ground
because there is no way for a golang application to fork a child exit
out of the parent process because all the golang threads stay with the
parent.
This patch has three parts that happen ONLY when $SIGUSR1_PARENT_PID
is set.
1) At the point where runc start would normally exit, it closes
stdin/stdout/stderr so it would be possible to daemonize "runc start ...".
2) The code to send a SIGUSR1 to the parent process was added. The
idea being that a parent process would simply exit at that point
because it was blocking until runc performed everything it was
required to perform.
3) The code was copied which performs the normal the signal handling
block which is used for the foreground operation of runc.
-- More information --
When you use "runc run " it is running in the "foreground", in the
sense it takes over your existing terminal.
The runc-docker doesn't have a way to start it with "runc run&" where
you can send it to the background and have everything work. With this
commit, it does allow you to do that and have all the stop hooks fire
at the time what ever runc started exits.
Lets take a quick look at what "runc run" does today:
* Starts a whole pile of threads
* Sets up all name spaces
* Starts child process for container and leaves it paused at image activation
* runs start hooks
* executes "continue" for container process
* waits for container app to exit
* executes stop hooks
Now lets look at "runc create/start" does today:
runc create
* Starts a whole pile of threads
* Sets up all name spaces
* Starts child process for container and leaves it paused at image activation
* exits -- [ NOTE: this is our problem! ]
runc start
* runs start hooks
* executes "continue" for continue process
At this point when the container app exits nothing is waiting for it
to run any kind of hooks.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The rdepends on glibc-utils was removed without reason
in the following commit:
"""
e73608d56e
lxc: 2.0.0 -> 2.0.8
"""
And it causes failure:
/usr/libexec/lxc/lxc-net: line 125: getent: command not found
So add the dependency back.
Signed-off-by: Jackie Huang <jackie.huang@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
These files were removed since the followine patch
http://xenbits.xen.org/gitweb/?p=xen.git;a=commit;h=976ce01e103fc03bb3336cdbc6c951a709e285fe
and are absent in xen 4.8.0 and further, which leads to
the following error during the Xen build process:
do_package: SYSTEMD_SERVICE_xen-xencommons value xenstored.socket does not exist
Signed-off-by: Yan Yankovskyi <yan.yankovskyi@globallogic.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Rather than throwing an error if we can't map a user to a uid,
output a warning. We aren't actually running the code, but are
just extracting it .. so the user not existing isn't an issue.
With this, we avoid the not-so-useful traces like this:
config.User: unsupported format
github.com/opencontainers/image-tools/image.(*config).runtimeSpec
oci-image-tools/0.2.0-dev+gitAUTOINC+4abe1a166f-r0/oci-image-tools-0.2.0-dev+gitAUTOINC+4abe1a16 6f/src/import/vendor/src/github.com/opencontainers/image-tools/image/config.go:109
<...>
src/import/cmd/oci-image-tool/main.go:57
runtime.main /usr/lib64/go/src/runtime/proc.go:185
runtime.goexit /usr/lib64/go/src/runtime/asm_amd64.s:2337
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Fix license path and checksum for python3-dockerpty.
Signed-off-by: Abdur Rehman <abdur_rehman@mentor.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Cleanup the various python-* recipes which download from pypi. The
biggest change is to "inherit pypi" which should result in us always
using current pypi best practices. This will for example ensure we are
using https and not http which is apparently going to be disabled
soon.
Where the default behavior in the pypi class wasn't sufficient we make
use of the PYPI_* variables to overwrite the defaults.
Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
To make reviews easier and to be consistent with other patch
submission guidelines in other layers, add the '-M' option to the 'git
send-email' line.
Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Fixed issues:
New features in this release:[From Neil Horman]
=================================
optimization of platform device irq detection
Added sample udev rules to trigger irq rescans on device add/remove
Made irqbalance ui an optional compile component
Added support for Intel CoD
Add -v | --version option to command line
Bug fixes in this release:
==================================
Misc compiler warning fixes and spelling errors
Compilation error fix on aarch64
Compilation error fix when using clang in c99 mode
Unused variable cleanup
Lots of memory leak cleanup in irqbalance-ui
Prevent irqbalance from running in a container
Fix irq affinity assignment in some cases to wrong numa node
Fix oneshot mode
exclude legacy irq 255
Deprecations:
=================================
without-glib2 is removed, we just need it enough that we can't dummy it up anymore
Tested:
================================
Tested on AARCH64 and it compiles
Signed-off-by: RC Reddy <vlsireddy@gmail.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
runv needs a kernel and initrd as part of its startup. The tested
configuration for these is hyperstart, so we add it to the rdepends
to ensure it is installed.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
For VMs masquerading as kernels, we need a kernel and initrd. runv
uses hyperstart for this purpose so we add the recipe here. Once
built, the kernel and initrd are installed to where runv can find
them automatically.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
If the root name space has additional cgroup mounts, pass them to the
container.
Signed-off-by: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The only user in meta-virtualization is python-bugsnag and there is no
version dependency. This new version is however required by the latest
openstack. Switched to using the "common" pypi.bbclass while we are at
it.
Signed-off-by: Mark Asselstine <mark.asselstine@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Fixed issues:
Usage of makedev requires including <sys/sysmacros.h> otherwise
the error is raised due to multiple definition in <sys/types.h>.
Add include path to kernel headers required to get kvmtool
compilatioin successful.
Signed-off-by: Dariusz Pelowski <dariusz.pelowski@gmail.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The variable that must be set to disable the virtualization distro
feature check was not documented in the README ... making it not
so useful.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Fix for building biossums on genericx86-64 machine, due to QA issue:
ERROR: QA Issue: No GNU_HASH in the elf binary:
'.../0.7a-r0/packages-split/biossums/usr/bin/biossums' [ldflags]
caused by overriding LDFLAGS variable to empty value in project's makefile.
In consequence there's missing LDFLAGS settings from bitbake (including
-Wl,--hash-style=gnu, which causes showing this QA).
Signed-off-by: Dariusz Pelowski <dariusz.pelowski@gmail.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Add python3-docker-pycreds, python3-requests and
python3-websocket-client to RDEPENDS for a functional python docker
module (otherwise even import docker fails).
Signed-off-by: Ricardo Salveti <ricardo@opensourcefoundries.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Ricardo Salveti <ricardo@opensourcefoundries.com> pointed out that
runc-docker was not getting a proper PV due to the use of SRCREV
in the variable.
By switching to SRCPV, we get the right PV for both variants of
runc.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The go bbclass already inherits goarch.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
This variable isn't picked up by the runc Makefile anyway as it isn't exported.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The "vendor/src" symlink is already created in do_compile in runc.inc.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The makefile for netns needs to be patched to ensure that we use the binaries
from go-cross and not go-native.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The recvtty demo/reference application has cross compilation issues when
targeting aarch64 platforms. As it is just a demo application and is not usually
used, we can just patch the Makefile to disable building this application.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The ConfigParsed event is raised multiple times during a build which resulted
in the sanity warning appearing up to 4 times per build. Instead we should be
watching for the SanityCheck event (this is what the sanity checks in oe-core
watch for).
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The variable used to skip the sanity check should be named in a layer specific
way. This will ensure there is no confusion if other layers use a similar sanity
check.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
If the GOARCH isn't set CGO took the ARCH from the host and tries
to use -m64 flag that isn't available in aarch64 compiler.
...
| # github.com/opencontainers/runc/libcontainer/system
| aarch64-linaro-linux-gcc: error: unrecognized command line option
'-m64'
| # github.com/containerd/console
| aarch64-linaro-linux-gcc: error: unrecognized command line option
'-m64'
| Makefile:60: recipe for target 'shim-static' failed
...
Signed-off-by: Aníbal Limón <anibal.limon@linaro.org>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
syslinux images cannot boot because they're missing of xen.gz
file. Current poky doesn't use populate() function, instead
syslinux_populate() can be used.
Signed-off-by: Dariusz Pelowski <dariusz.pelowski@gmail.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Many distros make a 'kubernetes-cni' package available. This is no
more than a binary extracted set of 'cni' plugins. We already have
'cni' available in the build system, so we make sure it rprovides
kubernetes-cni to keep conflicts and confusion at a minimum.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Introducing the kubernets components for container orchestration.
The packaging introduced here creates the framework for more
advanced kubernetes configurations. By itself, it doesn't do much
but it makes the components available to be configured and deployed
as master/worker nodes.
It integrates with the previously introduced cni and cri-o
components available in meta-virt.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
These variables are now set correctly by go.bbclass in oe-core. Changing them to
point at the native sysroot just leads to build errors in some cases, for
example when the target and host have matching GOARCH but not matching c
libraries.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The makefiles for both providers of runc need to be patched in similar ways to
ensure that we use the binaries from go-cross and not go-native.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
This patch hasn't been used in a long time.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
There's a problem with fetching SeaBIOS source code via http.
SeaBIOS server returns 308: Permanent Redirect which fails on
wget 1.17.1. The official source code uri goes to https, this
patch aligns SRC_URI to officialy provided by SeaBIOS team.
Signed-off-by: Dariusz Pelowski <dariusz.pelowski@gmail.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Remove recipe lines modifying DISTRO_FEATURES that were intended to
simplify the non-x86 x11 image dependencies, but did not.
Signed-off-by: Christopher Clark <christopher.clark6@baesystems.com>
Reported-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The logic for restricting the xen-pciback kernel module to x86-only
images was incorrect, resulting in the IMAGE_INSTALL_x86 and _x86-64
variables causing the IMAGE_INSTALL variable contents to be
overridden and incomplete.
This correction removes the specialized IMAGE_INSTALL_x86 and _x84-64
variables. The replacement logic causes a value with
conditionally-populated contents to be added to the IMAGE_INSTALL
variable.
Signed-off-by: Christopher Clark <christopher.clark6@baesystems.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
These fixes are needed due to updates to go.bbclass in oe-core. See commit
01a8d45370 for more information.
Signed-off-by: Paul Barker <pbarker@toganlabs.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Update README file to include information about bbappend file inclusion,
telling the users that 'virtualization' needs to be in DISTRO_FEATURES
to make some bbappend files to be effective.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Make these bbappend files to take effect only when DISTRO_FEATURES
contain 'virtualization'. Otherwise, we would meet failure failure
at system booting up qemux86. Related logs are as below.
systemd-modules-load[113]: Failed to insert 'kvm_amd': Operation not supported
systemd-modules-load[113]: Failed to insert 'kvm_intel': Operation not supported
Also, make kvm related settings depend on the 'kvm' distro feature.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Add a new class, sanity-meta-virt.bbclass, to check for whether necessary
settings are available for bbappend files in this layer to be effective,
and warn users if not.
In addition, a variable SKIP_SANITY_BBAPPEND_CHECK is added to enable users
to explicitly skip the checking to avoid unwanted warnings.
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Introducing runv, which is a Hypervisor-based Runtime for OCI.
runV is compatible with OCI. However, due to the difference between
hypervisors and containers, the following sections of OCI don't apply to runV:
- Namespace
- Capability
- Device
- linux and mount fields in OCI specs are ignored
- Hypervisor
The current release of runV supports the following hypervisors:
- KVM (QEMU 2.1 or later)
- KVM (Kvmtool)
- Xen (4.5 or later)
- QEMU without KVM (NOT RECOMMENDED. QEMU 2.1 or later)
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Introducing the CNI framework to allow platforms such as
kubernetes + cri-o to use thm for networking configuration.
The networking plugins are installed, but are not configured
by default to create any specific networking interfaces, which
are the responsibility of the image and distribution.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
To prepare for native kubernetes support without docker on a target,
we integrate the cri-o incubator project.
cri-o is meant to provide an integration path between OCI conformant
runtimes and the kubelet. Specifically, it implements the Kubelet
Container Runtime Interface (CRI) using OCI conformant runtimes.
The scope of cri-o is tied to the scope of the CRI.
This initial introduction is build + packaging only. It is expected
that configuration and deployment tweaks are done at the distro
level.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
The following commit:
commit df3a46feb971386f922c7c2c2822b88301f87cb0
Author: Ma Shimiao <mashimiao.fnst@cn.fujitsu.com>
Date: Tue Aug 1 17:39:39 2017 +0800
implement add/set function for hooks items
Signed-off-by: Ma Shimiao <mashimiao.fnst@cn.fujitsu.com>
Breaks the ability to specify multiple hooks with the same path
(i.e. a shell script that does different things based on arguments).
The author's intent with the change is unclear, so rather than fix
it, we revert it for now.
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
Fix the image for non-x86, both with and without x11 included.
* Only install xf86-video-vesa on x86 platforms as it is not
available elsewhere.
* In the absence of xf86-video-vesa (ie. on non-x86 platforms)
enable opengl and remove wayland from IMAGE_INSTALL to
satisfy build requirements for x11.
* Use IMAGE_FEATURES instead of DISTRO_FEATURES to enable x11
inclusion in the image.
* Verify the required DISTRO_FEATURE of 'xen' and if enabled, 'x11'.
* When building the Xserver with Xen in DISTRO_FEATURES, on non-x86,
turn off glamor by default, to remove dependency on egl.
Introduces: xserver-xorg_%.bbappend
Signed-off-by: Christopher Clark <christopher.clark6@baesystems.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
This fixes the build when libc-locale-code is not a DISTRO_FEATURE.
xen's configure script checks for xgettext and will abort if not found.
xgettext is provided by the gettext-native package but not by
gettext-minimal-native. This means that "inherit gettext" is not
sufficient to satisfy Xen's build requirements: gettext-native is
required.
Signed-off-by: Christopher Clark <christopher.clark6@baesystems.com>
Signed-off-by: Bruce Ashfield <bruce.ashfield@windriver.com>
