Skip to content
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

stable: Update to 6.12 #329

Closed
wants to merge 15 commits into from
Closed

stable: Update to 6.12 #329

wants to merge 15 commits into from

Conversation

1Naim
Copy link
Member

@1Naim 1Naim commented Nov 15, 2024

6.12 will be released in approx. 3 days, this is the branch that will prepare our PKGBUILDs for that kernel version, as it seems that 6.12 is quite a big release.

Changes in patchset from 6.11 to 6.12:

  • Updated AMD V-Cache Driver
  • crypto patch that speeds up workloads dealing with crc32 and AES128
  • The everchanging fixes patch

Honorable Mentions:

  • openvpn-dco - Part of 6.12 patchset until rc7. Module is unused by userspace so there is no reason for us to ship it.

PKGBUILD Changes:

  • Updated descriptions and comments for various kernels
  • Dropped linux-cachyos-sched-ext - sched-ext is upstream now!
  • Ported over changes made in linux-cachyos-rc/PKGBUILD to the stable kernel PKGBUILDs

Some considerations to be made:

  • linux-cachyos-rt-bore: Use PREEMPT_RT instead of PREEMPT_AUTO - We got no choice, PREEMPT_AUTO isn't a thing anymore

eevdf) ## 6.12 EEVDF patches
source+=("${_patchsource}/sched/0001-eevdf-next.patch");;
rt) ## EEVDF with RT patches
source+=("${_patchsource}/misc/0001-rt.patch");;
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why RT patch was removed?

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It should be not required anymore, since it got enabled in upstream. If we want to have latest RT features, from the rt tree, then we can keep it, but it also brings more maintenance work

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The commit of removing RT dependency on CONFIG_EXPERT can also be cherrypicked to, e.g cachy.patch

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

https://git.kernel.org/pub/scm/linux/kernel/git/rt/linux-rt-devel.git/log/?h=linux-6.12.y-rt-rebase This is the latest tree of the rt patchset. Initial thoughts are patch 1-3, 6-9 seem nice to have

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It should be not required anymore, since it got enabled in upstream. If we want to have latest RT features, from the rt tree, then we can keep it, but it also brings more maintenance work

From the remaining patches, we don't need the ARM/PowerPC/RISCV specific ones too much since we're not targeting them, but PREEMPT_LAZY and i915 patches are also here. The first one we'll want to have anyway for all kernels, but the second one can be separated into a single patch.

https://gitlab.archlinux.org/archlinux/packaging/upstream/linux-rt/-/tree/linux-6.12.y-rt-patches/patches

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

https://github.com/CachyOS/linux/commits/6.12/rt I've cherrypicked some patches from the rt patchset. This seems to meet the standard expectations of being part of our base patchset. The first two commits is also beneficial for normal kernels, and splitting these into different patches makes no sense in my eyes. The other commits are relatively safe to include for non-RT kernels.

@ptr1337 what do you think?

linux-cachyos-autofdo/PKGBUILD Show resolved Hide resolved
linux-cachyos-server/PKGBUILD Show resolved Hide resolved
@1Naim 1Naim marked this pull request as ready for review November 18, 2024 03:30
@1Naim
Copy link
Member Author

1Naim commented Nov 18, 2024

linux-cachyos and linux-cachyos-rt-bore has been compile-tested. Binaries can be found in https://share.cachyos.org/x86_64_v3/kernel/stable-testing/. Patches used are based on https://github.com/CachyOS/linux/commits/6.12/base-dev. I'll try to rebase this PR to fix the merge conflicts once the tarballs are available in https://cdn.kernel.org.

1Naim and others added 11 commits November 18, 2024 15:07
sched-ext has finally been merged in 6.12. There is no longer any reason to use this kernel. For comparisons between cachy's
patched EEVDF, consider `linux-cachyos-eevdf`. For stock EEVDF, consider `linux-cachyos-server`.

Signed-off-by: Eric Naim <[email protected]>
Sync changes from `linux-cachyos-rc/PKGBUILD` to all stable kernel PKGBUILDs where applicable.

Signed-off-by: Eric Naim <[email protected]>
This should differentiate it from linux-cachyos-bore

Signed-off-by: Eric Naim <[email protected]>
Minor tweaks to the switch cases and comments.

Signed-off-by: Eric Naim <[email protected]>
Minor tweaks to the switch cases and comments.

Signed-off-by: Eric Naim <[email protected]>
PREEMPT_AUTO was part of the out of tree realtime patchset and isn't included in the mainline kernel. The latest revision
of the realtime patchset has also dropped PREEMPT_AUTO (it lacked documentation, I'm not sure what it's really for).

Signed-off-by: Eric Naim <[email protected]>
We don't need `bore-cachy-rt` anymore. This can be reverted if we want to pull out of tree rt patches *only* for the rt kernel,
however I much prefer including the rt patchset thats from https://github.com/CachyOS/linux/tree/6.12/rt as one of
our base patchsets instead.

Signed-off-by: Eric Naim <[email protected]>
Signed-off-by: Eric Naim <[email protected]>
Copy link
Member

@sirlucjan sirlucjan left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It looks OK to me, but with one comment: maybe we would leave the sched-ext kernel and add a patch to it https://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git/ because I think @htejun might be interested in that.

Copy link
Member

@ventureoo ventureoo left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

diff --git a/linux-cachyos/PKGBUILD b/linux-cachyos/PKGBUILD
index 3ee2b5ff..6e01b3b3 100644
--- a/linux-cachyos/PKGBUILD
+++ b/linux-cachyos/PKGBUILD
@@ -242,13 +242,12 @@ fi
 
 ## List of CachyOS schedulers
 case "$_cpusched" in
-    cachyos|bore|rt-bore) # CachyOS Scheduler (BORE)
-        source+=("${_patchsource}/sched/0001-bore-cachy.patch");;
+    cachyos|bore|rt-bore|hardened) # CachyOS Scheduler (BORE)
+        source+=("${_patchsource}/sched/0001-bore-cachy.patch");;&
     bmq) ## Project C Scheduler
         source+=("${_patchsource}/sched/0001-prjc-cachy.patch");;
-    hardened) ## Hardened Patches with BORE Scheduler
-        source+=("${_patchsource}/sched/0001-bore-cachy.patch"
-                 "${_patchsource}/misc/0001-hardened.patch");;
+    hardened) ## Hardened Patches
+        source+=("${_patchsource}/misc/0001-hardened.patch");;
 esac
 
 export KBUILD_BUILD_HOST=cachyos

This makes it possible to merge `hardened` with the other options that source BORE (cachyos, bore, rt-bore) while also
keeping a separate pattern that sources the hardened patchset.

Signed-off-by: Eric Naim <[email protected]>
@1Naim 1Naim requested a review from ventureoo November 18, 2024 11:30
Copy link
Member

@ventureoo ventureoo left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

CI failed, because .SRCINFO mismatched, but otherwise LGTM.

@1Naim
Copy link
Member Author

1Naim commented Nov 18, 2024

CI failed, because .SRCINFO mismatched, but otherwise LGTM.

Yep, was planning on updating .SRCINFO when Pete synced the patches. All the individual trees should merge fine on a clean 6.12 tree, but IDK which patches are going to be included and which are not, so I'd rather wait.

edit: It would have failed even if .SRCINFO was updated because the base patch doesn't apply.

1Naim and others added 2 commits November 18, 2024 20:43
@ptr1337
Copy link
Member

ptr1337 commented Nov 18, 2024

Pushed with 503f483

@ptr1337 ptr1337 closed this Nov 18, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants