I already have version bumps for all of the related components under
test, they'll show up in master-next shortly, and should cover this.

Cheers,

Bruce

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

Bump from v1.4.4 to v.1.4.12 so that some CVEs are resolved,
e.g. CVE-2021-41103.

Signed-off-by: Chen Qi <Qi.Chen@...>
---
.../containerd/containerd-opencontainers_git.bb | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/recipes-containers/containerd/containerd-opencontainers_git.bb b/recipes-containers/containerd/containerd-opencontainers_git.bb
index 774a28c..7f6c75d 100644
--- a/recipes-containers/containerd/containerd-opencontainers_git.bb
+++ b/recipes-containers/containerd/containerd-opencontainers_git.bb
@@ -5,7 +5,7 @@ DESCRIPTION = "containerd is a daemon to control runC, built for performance and
support as well as checkpoint and restore for cloning and live migration of containers."


-SRCREV = "409c87ba59dd96965239573aa9458a3585c05468"
+SRCREV = "7b11cfaabd73bb80907dd23182b9347b4245eb5d"
SRC_URI = "git://github.com/containerd/containerd;branch=release/1.4 \
file://0001-build-use-oe-provided-GO-and-flags.patch \
file://0001-Add-build-option-GODEBUG-1.patch \
@@ -15,7 +15,7 @@ SRC_URI = "git://github.com/containerd/containerd;branch=release/1.4 \
LICENSE = "Apache-2.0"
LIC_FILES_CHKSUM = "file://src/import/LICENSE;md5=1269f40c0d099c21a871163984590d89"

-CONTAINERD_VERSION = "v1.4.4"
+CONTAINERD_VERSION = "v1.4.12"

EXTRA_OEMAKE += "GODEBUG=1"

--
2.33.0

Signed-off-by: Yi Zhao <yi.zhao@...>
---
meta-openstack/Documentation/README.OpenLDAP | 2 +-
meta-openstack/README.setup | 2 +-
2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/meta-openstack/Documentation/README.OpenLDAP b/meta-openstack/Documentation/README.OpenLDAP
index a45b769..95c2227 100644
--- a/meta-openstack/Documentation/README.OpenLDAP
+++ b/meta-openstack/Documentation/README.OpenLDAP
@@ -6,7 +6,7 @@ OpenLDAP into DISTRO_FEATURES

e.g. in conf/local.conf

-DISTRO_FEATURES:append += " OpenLDAP"
+DISTRO_FEATURES:append = " OpenLDAP"

A number of variables can be specified during the build phase that configures
OpenLDAP specific options:
diff --git a/meta-openstack/README.setup b/meta-openstack/README.setup
index d1a4703..f837f63 100644
--- a/meta-openstack/README.setup
+++ b/meta-openstack/README.setup
@@ -95,7 +95,7 @@ systemd will be used in your images:

Additionally activiate the meta-virtualization layer:

- DISTRO_FEATURES:append += "virtualization kvm"
+ DISTRO_FEATURES:append = " virtualization kvm"


Package configurations
--
2.25.1

Fixes:
WARNING: openstack-image-compute.bb: IMAGE_ROOTFS_EXTRA_SPACE:append +=
is not a recommended operator combination, please replace it.

Signed-off-by: Yi Zhao <yi.zhao@...>
---
.../recipes-extended/images/openstack-image-compute.bb | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-openstack/recipes-extended/images/openstack-image-compute.bb b/meta-openstack/recipes-extended/images/openstack-image-compute.bb
index 15726fd..a725e2e 100644
--- a/meta-openstack/recipes-extended/images/openstack-image-compute.bb
+++ b/meta-openstack/recipes-extended/images/openstack-image-compute.bb
@@ -19,6 +19,6 @@ inherit monitor

# Ensure extra space for guest images, and rabbit MQ has a hard coded
# check for 2G of free space, so we use 3G as a starting point.
-IMAGE_ROOTFS_EXTRA_SPACE:append += "+ 3000000"
+IMAGE_ROOTFS_EXTRA_SPACE:append = " + 3000000"

# ROOTFS_POSTPROCESS_COMMAND += "remove_packaging_data_files ; "
--
2.25.1

Fixes:
WARNING: openstack-image-controller.bb: IMAGE_ROOTFS_EXTRA_SPACE:append
+= is not a recommended operator combination, please replace it.

Signed-off-by: Yi Zhao <yi.zhao@...>
---
.../recipes-extended/images/openstack-image-controller.bb | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-openstack/recipes-extended/images/openstack-image-controller.bb b/meta-openstack/recipes-extended/images/openstack-image-controller.bb
index 7f41a08..96dc80c 100644
--- a/meta-openstack/recipes-extended/images/openstack-image-controller.bb
+++ b/meta-openstack/recipes-extended/images/openstack-image-controller.bb
@@ -24,5 +24,5 @@ inherit monitor

# Ensure extra space for guest images, and rabbit MQ has a hard coded
# check for 2G of free space, so we use 5G as a starting point.
-IMAGE_ROOTFS_EXTRA_SPACE:append += "+ 5000000"
+IMAGE_ROOTFS_EXTRA_SPACE:append = " + 5000000"

--
2.25.1

On Fri, Nov 19, 2021 at 8:43 AM kai <kai.kang@...> wrote:

On 11/11/21 7:05 PM, Peter Kjellerstedt wrote:

-----Original Message-----
From: meta-virtualization@... <meta-
virtualization@...> On Behalf Of kai
Sent: den 11 november 2021 11:07
To: meta-virtualization@...
Subject: [meta-virtualization][m-c-s][PATCH] concurrent-ruby: add explicit
branch

From: Kai Kang <kai.kang@...>

Add explicit branch master for concurrent-ruby as the same for other
recipes.

Signed-off-by: Kai Kang <kai.kang@...>
---
meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
b/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
index a328203e..61a2c598 100644
--- a/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
+++ b/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
@@ -4,7 +4,7 @@ HOMEPAGE = "http://www.concurrent-ruby.com"
LICENSE = "MIT"
LIC_FILES_CHKSUM =
"file://LICENSE.md;md5=fde65ae93d18826f70c6fe125aa04297"

-SRC_URI = "git://github.com/ruby-concurrency/concurrent-ruby.git;protocol=https;tag=v1.1.6\
+SRC_URI = "git://github.com/ruby-concurrency/concurrent-ruby.git;protocol=https;tag=v1.1.6;branch=master \

While at it, why not remove the tag= and replace it with a proper
SRCREV = "<sha1>"?

Sorry for late reply.

I am not sure which is better because there is a tag with branch in
SRC_URI in meta-oe. If you insist remove tag is better, I'll send it
with v2.

The default is to use SRCREVs instead of tags, since tags can change
(although they shouldn't), and it used to be that tags did trigger
some network access with some configurations.

So if it isn't too much trouble, I do think switching to a SRCREV is better.

Bruce

Thanks,

Kai

file://0001-Removed-check-for-concurrent_ruby.jar.patch"

S = "${WORKDIR}/git"
--
2.17.1

//Peter

--
Kai Kang
Wind River Linux

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

On 11/11/21 7:05 PM, Peter Kjellerstedt wrote:

-----Original Message-----
From: meta-virtualization@... <meta-
virtualization@...> On Behalf Of kai
Sent: den 11 november 2021 11:07
To: meta-virtualization@...
Subject: [meta-virtualization][m-c-s][PATCH] concurrent-ruby: add explicit
branch

From: Kai Kang <kai.kang@...>

Add explicit branch master for concurrent-ruby as the same for other
recipes.

Signed-off-by: Kai Kang <kai.kang@...>
---
meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
b/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
index a328203e..61a2c598 100644
--- a/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
+++ b/meta-openstack/recipes-devtools/ruby/concurrent-ruby_1.1.6.bb
@@ -4,7 +4,7 @@ HOMEPAGE = "http://www.concurrent-ruby.com"
LICENSE = "MIT"
LIC_FILES_CHKSUM =
"file://LICENSE.md;md5=fde65ae93d18826f70c6fe125aa04297"

-SRC_URI = "git://github.com/ruby-concurrency/concurrent-ruby.git;protocol=https;tag=v1.1.6\
+SRC_URI = "git://github.com/ruby-concurrency/concurrent-ruby.git;protocol=https;tag=v1.1.6;branch=master \

While at it, why not remove the tag= and replace it with a proper
SRCREV = "<sha1>"?

Sorry for late reply.

I am not sure which is better because there is a tag with branch in SRC_URI in meta-oe. If you insist remove tag is better, I'll send it with v2.

Thanks,

Kai

file://0001-Removed-check-for-concurrent_ruby.jar.patch"

S = "${WORKDIR}/git"
--
2.17.1

//Peter

--
Kai Kang
Wind River Linux

merged.

Bruce

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

merged.

Bruce

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

On Fri, Nov 19, 2021 at 2:10 AM kai <kai.kang@...> wrote:

On 11/11/21 11:53 AM, kai wrote:

From: Kai Kang <kai.kang@...>

There are typo spaces in SRC_URI which are introduced when add
';branch=master'. Remove them.

Signed-off-by: Kai Kang <kai.kang@...>
---
.../recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb b/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb
index d4c489cf..fe802b72 100644
--- a/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb
+++ b/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb
@@ -9,7 +9,7 @@ LIC_FILES_CHKSUM = "file://COPYING;md5=b234ee4d69f5fce4486a80fdaf4a4263"
DEPENDS = "kmod openssl util-linux open-isns"
DEPENDS:append = " ${@bb.utils.contains('DISTRO_FEATURES', 'systemd', 'systemd', '', d)}"

-SRC_URI = "git://github.com/open-iscsi/open-iscsi.git;protocol=https ;branch=master \
+SRC_URI = "git://github.com/open-iscsi/open-iscsi.git;protocol=https;branch=master \
file://0001-fix-build-error-of-cross-build.patch \
"


Ping.

I was waiting for you to reply about the comment on your other patch
in the same send (the ruby branch one), since if they are sent on the
same day by the same person, and there are comments .. I tend to treat
them as a unit.

But I have gone ahead and merged this now.

Bruce

--
Kai Kang
Wind River Linux

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

From: Kai Kang <kai.kang@...>

Fix warning of openstack-image-aio:

| WARNING: /path/to/meta-openstack/recipes-extended/images/openstack-image-aio.bb:
| IMAGE_ROOTFS_EXTRA_SPACE:append += is not a recommended operator combination,
| please replace it.

Signed-off-by: Kai Kang <kai.kang@...>
---
meta-openstack/recipes-extended/images/openstack-image-aio.bb | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-openstack/recipes-extended/images/openstack-image-aio.bb b/meta-openstack/recipes-extended/images/openstack-image-aio.bb
index 858c8e99..f01d0ed5 100644
--- a/meta-openstack/recipes-extended/images/openstack-image-aio.bb
+++ b/meta-openstack/recipes-extended/images/openstack-image-aio.bb
@@ -27,7 +27,7 @@ inherit identity
inherit monitor

# check for 5G of free space, so we use 5G as a starting point.
-IMAGE_ROOTFS_EXTRA_SPACE:append += "+ 5000000"
+IMAGE_ROOTFS_EXTRA_SPACE += "+ 5000000"

POST_KEYSTONE_SETUP_COMMAND = "/etc/keystone/hybrid-backend-setup"

--
2.17.1

From: Kai Kang <kai.kang@...>

There are typo spaces in SRC_URI which are introduced when add
';branch=master'. Remove them.

Signed-off-by: Kai Kang <kai.kang@...>
---
 .../recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb    | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb b/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb
index d4c489cf..fe802b72 100644
--- a/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb
+++ b/meta-openstack/recipes-connectivity/open-iscsi/open-iscsi-user_2.1.4.bb
@@ -9,7 +9,7 @@ LIC_FILES_CHKSUM = "file://COPYING;md5=b234ee4d69f5fce4486a80fdaf4a4263"
 DEPENDS = "kmod openssl util-linux open-isns"
 DEPENDS:append = " ${@bb.utils.contains('DISTRO_FEATURES', 'systemd', 'systemd', '', d)}"
 
-SRC_URI = "git://github.com/open-iscsi/open-iscsi.git;protocol=https  ;branch=master \
+SRC_URI = "git://github.com/open-iscsi/open-iscsi.git;protocol=https;branch=master \
            file://0001-fix-build-error-of-cross-build.patch \
            "
 

Ping.

-- 
Kai Kang
Wind River Linux

When enabling 'dpdk' PACKAGECONFIG, the following error appears.

| configure: error: Could not find DPDK library in default search path

Fix the error by tweaking the configure option regarding dpdk.
Add pkgconfig to 'inherit' because pkgconfig is required at do_configure
when dpdk is enabled.

Signed-off-by: Chen Qi <Qi.Chen@...>
---
recipes-networking/openvswitch/openvswitch.inc | 2 +-
recipes-networking/openvswitch/openvswitch_git.bb | 4 +---
2 files changed, 2 insertions(+), 4 deletions(-)

diff --git a/recipes-networking/openvswitch/openvswitch.inc b/recipes-networking/openvswitch/openvswitch.inc
index b354a84..831e0b3 100644
--- a/recipes-networking/openvswitch/openvswitch.inc
+++ b/recipes-networking/openvswitch/openvswitch.inc
@@ -72,7 +72,7 @@ FILES:${PN} += "${datadir}/ovsdbmonitor"
FILES:${PN} += "/run"

FILES:${PN} += "${libdir}/python${PYTHON_BASEVERSION}/"
-inherit autotools update-rc.d systemd python3native
+inherit pkgconfig autotools update-rc.d systemd python3native

SYSTEMD_PACKAGES = "${PN}-switch"
SYSTEMD_SERVICE:${PN}-switch = " \
diff --git a/recipes-networking/openvswitch/openvswitch_git.bb b/recipes-networking/openvswitch/openvswitch_git.bb
index f1a5c4c..0fb7c13 100644
--- a/recipes-networking/openvswitch/openvswitch_git.bb
+++ b/recipes-networking/openvswitch/openvswitch_git.bb
@@ -32,10 +32,8 @@ SRC_URI += "git://github.com/openvswitch/ovs.git;protocol=https;branch=branch-2.

LIC_FILES_CHKSUM = "file://LICENSE;md5=1ce5d23a6429dff345518758f13aaeab"

-DPDK_INSTALL_DIR ?= "/opt/dpdk"
-
PACKAGECONFIG ?= "libcap-ng"
-PACKAGECONFIG[dpdk] = "--with-dpdk=${STAGING_DIR_TARGET}${DPDK_INSTALL_DIR}/share/${TARGET_ARCH}-native-linuxapp-gcc,,dpdk,dpdk"
+PACKAGECONFIG[dpdk] = "--with-dpdk=shared,,dpdk,dpdk"
PACKAGECONFIG[libcap-ng] = "--enable-libcapng,--disable-libcapng,libcap-ng,"
PACKAGECONFIG[ssl] = ",--disable-ssl,openssl,"

--
2.33.0

No worries, now it's my turn to apologize :) Got pulled away from this over the last week and haven't had a chance to fully dive back on in, but I really appreciate you taking the time to reply here! Now that I've had more time with Yocto and the meta-virt layer I think I'm closing the knowledge gaps that are blocking me, I just have a few more edges to sort out. I think that you're right that a patch shouldn't be needed, I think it's just been a lot of user error on my part.

I'll be sure to update this thread once I have the device up and running; I think I'm close. Thanks again!

On Thu, Nov 11, 2021 at 11:49 PM Kai <kai.kang@...> wrote:

On 11/12/21 12:03 PM, Bruce Ashfield wrote:

On Thu, Nov 11, 2021 at 10:23 PM kai <kai.kang@...> wrote:

From: Kai Kang <kai.kang@...>

/var/log is normally a link to /var/volatile/log and /var/volatile is a tmpfs
mount. So anything created in /var/log will not be available when the tmpfs is
mounted.

Not install /var/log to avoid 'empty-dirs' qa issue:

| ERROR: libvirt-7.2.0-r0 do_package_qa: QA Issue: libvirt installs files in
/var/volatile, but it is expected to be empty [empty-dirs]

Signed-off-by: Kai Kang <kai.kang@...>
---
recipes-extended/libvirt/libvirt_7.2.0.bb | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/recipes-extended/libvirt/libvirt_7.2.0.bb b/recipes-extended/libvirt/libvirt_7.2.0.bb
index b7f8383a..b49bd11a 100644
--- a/recipes-extended/libvirt/libvirt_7.2.0.bb
+++ b/recipes-extended/libvirt/libvirt_7.2.0.bb
@@ -232,6 +232,10 @@ do_install:append() {
echo "d root root 0755 /run/libvirt/qemu none" \
>> ${D}${sysconfdir}/default/volatiles/99_libvirt

+ # /var/log is normally a link to /var/volatile/log
+ # and /var/volatile is a tmpfs mount which should be clean.
+ rm -rf ${D}${localstatedir}/log

We should be preventing the install from happening at all, not
deleting it after the fact.

If libvirt thinks something needs to be in the directory, then we
should be looking into why it does exactly that.

OK.I'll figure out.

I poked around at this myself, and then asked Ross to help me out .. I
assumed it was meson doing this, and he pointed out the right
meson.build files to me.

I'm trying to figure out how to balance some non volatile use cases,
with this new QA check. If we universally delete those files, then
those cases are broken.

I'm leaning towards leaving the directories in place, and just
inhibiting the QA check.

Bruce

Regards,
Kai

Bruce

+
# Manually set permissions and ownership to match polkit recipe
if ${@bb.utils.contains('PACKAGECONFIG', 'polkit', 'true', 'false', d)}; then
install -d -m 0700 ${D}/${datadir}/polkit-1/rules.d
--
2.33.0

--
Kai Kang
Wind River Linux

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

On 11/12/21 12:03 PM, Bruce Ashfield wrote:

On Thu, Nov 11, 2021 at 10:23 PM kai <kai.kang@...> wrote:

From: Kai Kang <kai.kang@...>

/var/log is normally a link to /var/volatile/log and /var/volatile is a tmpfs
mount. So anything created in /var/log will not be available when the tmpfs is
mounted.

Not install /var/log to avoid 'empty-dirs' qa issue:

| ERROR: libvirt-7.2.0-r0 do_package_qa: QA Issue: libvirt installs files in
/var/volatile, but it is expected to be empty [empty-dirs]

Signed-off-by: Kai Kang <kai.kang@...>
---
recipes-extended/libvirt/libvirt_7.2.0.bb | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/recipes-extended/libvirt/libvirt_7.2.0.bb b/recipes-extended/libvirt/libvirt_7.2.0.bb
index b7f8383a..b49bd11a 100644
--- a/recipes-extended/libvirt/libvirt_7.2.0.bb
+++ b/recipes-extended/libvirt/libvirt_7.2.0.bb
@@ -232,6 +232,10 @@ do_install:append() {
echo "d root root 0755 /run/libvirt/qemu none" \
>> ${D}${sysconfdir}/default/volatiles/99_libvirt

+ # /var/log is normally a link to /var/volatile/log
+ # and /var/volatile is a tmpfs mount which should be clean.
+ rm -rf ${D}${localstatedir}/log

We should be preventing the install from happening at all, not
deleting it after the fact.

If libvirt thinks something needs to be in the directory, then we
should be looking into why it does exactly that.

OK.I'll figure out.

Regards,
Kai

Bruce

+
# Manually set permissions and ownership to match polkit recipe
if ${@bb.utils.contains('PACKAGECONFIG', 'polkit', 'true', 'false', d)}; then
install -d -m 0700 ${D}/${datadir}/polkit-1/rules.d
--
2.33.0

--
Kai Kang
Wind River Linux

On Thu, Nov 11, 2021 at 10:23 PM kai <kai.kang@...> wrote:

From: Kai Kang <kai.kang@...>

/var/log is normally a link to /var/volatile/log and /var/volatile is a tmpfs
mount. So anything created in /var/log will not be available when the tmpfs is
mounted.

Not install /var/log to avoid 'empty-dirs' qa issue:

| ERROR: libvirt-7.2.0-r0 do_package_qa: QA Issue: libvirt installs files in
/var/volatile, but it is expected to be empty [empty-dirs]

Signed-off-by: Kai Kang <kai.kang@...>
---
recipes-extended/libvirt/libvirt_7.2.0.bb | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/recipes-extended/libvirt/libvirt_7.2.0.bb b/recipes-extended/libvirt/libvirt_7.2.0.bb
index b7f8383a..b49bd11a 100644
--- a/recipes-extended/libvirt/libvirt_7.2.0.bb
+++ b/recipes-extended/libvirt/libvirt_7.2.0.bb
@@ -232,6 +232,10 @@ do_install:append() {
echo "d root root 0755 /run/libvirt/qemu none" \
>> ${D}${sysconfdir}/default/volatiles/99_libvirt

+ # /var/log is normally a link to /var/volatile/log
+ # and /var/volatile is a tmpfs mount which should be clean.
+ rm -rf ${D}${localstatedir}/log

We should be preventing the install from happening at all, not
deleting it after the fact.

If libvirt thinks something needs to be in the directory, then we
should be looking into why it does exactly that.

Bruce

+
# Manually set permissions and ownership to match polkit recipe
if ${@bb.utils.contains('PACKAGECONFIG', 'polkit', 'true', 'false', d)}; then
install -d -m 0700 ${D}/${datadir}/polkit-1/rules.d
--
2.33.0

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

From: Kai Kang <kai.kang@...>

/var/log is normally a link to /var/volatile/log and /var/volatile is a tmpfs
mount. So anything created in /var/log will not be available when the tmpfs is
mounted.

Not install /var/log to avoid 'empty-dirs' qa issue:

| ERROR: libvirt-7.2.0-r0 do_package_qa: QA Issue: libvirt installs files in
/var/volatile, but it is expected to be empty [empty-dirs]

Signed-off-by: Kai Kang <kai.kang@...>
---
recipes-extended/libvirt/libvirt_7.2.0.bb | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/recipes-extended/libvirt/libvirt_7.2.0.bb b/recipes-extended/libvirt/libvirt_7.2.0.bb
index b7f8383a..b49bd11a 100644
--- a/recipes-extended/libvirt/libvirt_7.2.0.bb
+++ b/recipes-extended/libvirt/libvirt_7.2.0.bb
@@ -232,6 +232,10 @@ do_install:append() {
echo "d root root 0755 /run/libvirt/qemu none" \
>> ${D}${sysconfdir}/default/volatiles/99_libvirt

+ # /var/log is normally a link to /var/volatile/log
+ # and /var/volatile is a tmpfs mount which should be clean.
+ rm -rf ${D}${localstatedir}/log
+
# Manually set permissions and ownership to match polkit recipe
if ${@bb.utils.contains('PACKAGECONFIG', 'polkit', 'true', 'false', d)}; then
install -d -m 0700 ${D}/${datadir}/polkit-1/rules.d
--
2.33.0

Sorry for the slow reply,

On Fri, Nov 5, 2021 at 11:45 PM Patrick Godwin <pbg.dev@...> wrote:

I'm in the process of building xen-image-minimal for my Raspberry Pi 4 Compute Module and am hitting some issues when running the image on physical hardware. Upon inspecting the boot partition of the sd-card, I noticed that the image only contained the dtb for the Raspberry Pi 4B; this makes sense when I look at xen-raspberrypi4-64.inc:

Not sure if you've sorted this out yet.

We do have reference images that booted on the RPI, but I can't say that I've
been testing them myself. I'm hoping that my reply will catch the attention of
those that do have the h/w and they can comment in more detail.

# Override the meta-raspberrypi default kernel preference
PREFERRED_PROVIDER_virtual/kernel ?= "linux-yocto"
KERNEL_DEVICETREE ?= "broadcom/bcm2711-rpi-4-b.dtb"

Compared with the device tree blobs declared in meta-raspberrypi's raspberrypi4-64.conf:

RPI_KERNEL_DEVICETREE = " \
broadcom/bcm2711-rpi-4-b.dtb \
broadcom/bcm2711-rpi-400.dtb \
broadcom/bcm2711-rpi-cm4.dtb \
"

Looking into it further, it seems that the device tree listing in the kernel used in the xen-image-minimal build doesn't contain the newer firmware blobs contained in meta-raspberrypi, which causes devices like ethernet to fail to start up when the image finally boots.

There's a different level of support on the h/w between linux-yocto and the
'vendor' rpi tree. We use linux-yocto for the core enablement because the
branches are stable/not rebased and have a cadence we can predict (all
mentioned things are breakages we've hit before!).

That being said, you can change the kernel provider to the linux-rpi, and
the rest of the meta-virt settings, etc, are still applicable and should work.

I've tried adding the bcm2711-rpi-cm4 firmware blob to the SD card manually, but that causes u-boot to fail with the error "Bad Linux ARM64 Image Magic!" after the Boot Xen step in output. I also tried manually replacing the bootfiles written by the xen sd card image with the latest blobs from meta-raspberrypi's packages, but once the OS boots I find that xen is no longer running, making me suspect I screwed up the configuration somewhere.

It could be a kernel configuration issue between the two images, if
the Xen packages are on the image, but nothing is running. Can you
interact with the Xen support via the xen cli at all ? Just to get a
better error message ?

Is this something I can trivially fix? Is there a way for me to override the device tree selection used by the minimal xen image in my local.conf? Or do I need to investigate patching one of the meta-virtualization recipes?

Anything that needs changing, should be overridable via variable or
through bbappends, so hopefully no patching is required.

Bruce

(Apologies if I've used any of the wrong lingo/terms here; still new to RasPi/Yocto/Xen :D)

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II

My xen-image-minimal build worked with this change, so I've grabbed it and
pushed it to master.

Bruce

--
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II