-
Notifications
You must be signed in to change notification settings - Fork 162
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Bump up XEN version to 4.18.0 #3534
Conversation
It looks like it has support. |
oh, thanks for finding it! I'll give it a try. |
@@ -20,7 +20,7 @@ RUN [ -f "$(basename ${XEN_SOURCE})" ] || tar --absolute-names -xz < /xen.tar.gz | |||
WORKDIR /xen/xen | |||
COPY *.patch arch /tmp/ | |||
RUN cp /tmp/"$(uname -m)"/*.patch /tmp/ | |||
RUN for p in /tmp/*.patch ; do patch -p1 < "$p" || exit 1 ; done | |||
#RUN for p in /tmp/*.patch ; do patch -p1 < "$p" || exit 1 ; done |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should we instead delete the content of the patch directories? Or will we need some of them?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We still need them, but it's just a temporary workaround to avoid applying the not-yet-ported patches.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
(it should be converted to a draft PR)
@@ -108,8 +103,6 @@ const qemuConfTemplate = `# This file is automatically generated by domainmgr | |||
driver = "intel-iommu" | |||
caching-mode = "on" | |||
{{ end }} | |||
[realtime] |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
According to QEMU's documentation:
The -realtime mlock=on|off argument has been replaced by the -overcommit mem-lock=on|off argument.
Even if mem-lock=off is the default it can make sense to let this option explicit in the config... or at least explain the removal in your commit message
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Essentially, this option enables or disables mlock() call (https://www.man7.org/linux/man-pages/man2/mlock.2.html). This is probably workload-specific.
@@ -54,12 +54,7 @@ const qemuConfTemplate = `# This file is automatically generated by domainmgr | |||
[machine] | |||
type = "{{.Machine}}" | |||
dump-guest-core = "off" | |||
{{- if .DomainStatus.CPUs }} |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why are these being removed? I think it will affect the unit tests, so you need to update them as well (same for the realtime config).... the removal reason should also be explained in the commit message...
Edit: I see... you didn't port @OhmSpectator patches yet...
pkg/xen-tools/Dockerfile
Outdated
ENV PKGS_arm64 libfdt | ||
|
||
RUN eve-alpine-deploy.sh | ||
|
||
ADD https://github.com/ninja-build/ninja.git#release /ninja | ||
WORKDIR /ninja | ||
RUN python3 ./configure.py --bootstrap |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we really need all these layers? Is this build too heavy? I think you could squash all these commands into a single RUN....
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@rene Are you sure intermediate layers are stored? It really makes it less readable to have these long blocks of RUN
all connected with &&
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes @christoph-zededa , that's the definition of RUN command on docker... @rucoder , I understand that you need to build ninja, that's fine, my point is that you can just group the commands (using &&) instead of use multiple RUN commands and create one layer per command....
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@rene since docker is using buildkit I have not seen them.
just to clarify: this is a WIP and very DRAFT PR for experementing with latest XEN/QEMU so I can share my work with other people. I've updated the description |
The new QEMU version natively supports CPU Pinning options, making the CPU pinning patches unnecessary. Here is the branch that adapts Pillar to use these options: |
@deitch ninja-build is not ninja :) but |
480fd81
to
79482da
Compare
- I had to build ninja from source since Alpine dropped a support for it in 3.8 Highlights: - QEMU 8.0.4 - libgcc instead of musl Signed-off-by: Mikhail Malyshev <[email protected]>
Signed-off-by: Mikhail Malyshev <[email protected]>
79482da
to
729c4e2
Compare
@rucoder still plan to work on that? or can be closed? |
Closing this in a favor of already merged #4186 |
Highlights:
TODO:
git format-patch
and lack all required informationapk add ninja-build
does the trick instead of building ninja from source@eriknordmark Looking at XEN release page at https://downloads.xenproject.org/release/xen/ I would expect XEN 4.18.0 in December this year.