@ proxy-maintainers, please update the tree with the ebuild and metadata attached. Reproducible: Always
Created attachment 327676 [details] 3.6.2 release ebuild Ebuild for the 3.6.2 release. Honors a new useflag : kvm to be set in order to use the kernel on kvm.
Created attachment 327678 [details, diff] Patch designed to allow boot on kvm This patch will be committed if the kvm useflag is set and will enable to boot the kernel on kvm.
Created attachment 327680 [details] New package metadata Metadatas updated to describe the new "kvm" useflag.
Please don't CC people yourself. Let the bug-wranglers do their job ;)
Conditional patching? why? Can't we just apply it unconditionally? This means that the rest of the *-sources (pf, gentoo, vanilla) wont work on a kvm?
(In reply to comment #5) > Conditional patching? why? Can't we just apply it unconditionally? This > means that the rest of the *-sources (pf, gentoo, vanilla) wont work on a > kvm? A/ This just concerns BFS-ized kernels starting from 3.6 => gentoo and vanilla are not concerned. B/ It has been observed that these kernels can't boot on kvm for possibly different reasons among which one has been clearly isolated : The "culprit" is a sequence of code enabling BFS to operate in a "voluntary preempt" context when rescheduling. C/ The patch basically deletes this sequence of code. Hence, with the side effect of possibly breaking a CONFIG_PREEMPT_VOLUNTARILY=y configured kernel. => Applying this patch unconditionally would have the consequence to possibly break "voluntary preempt" for everybody. v.g : Even for the standard user who, of course, does not boot on kvm. I do not feel sensible to break standard user's configuration => a standard emerge with default settings will bring a kernel working whatever his config options. The kvm user is supposed to know what he is doing as well as the consequences of his choices in terms of potential limitations. Hence the justification of conditional patching.
Well I still don't like it but again, this is a kernel patchset that has no upstream support (upstream as in LKML) so I guess it is ok to treat it that way.
(In reply to comment #7) > Well I still don't like it. I never expected anyone to like the workaround of the side effects of a disliked workaround.
+*ck-sources-3.6.2 (29 Oct 2012) + + 29 Oct 2012; Markos Chandras <hwoarang@gentoo.org> +ck-sources-3.6.2.ebuild, + +files/ck-sources-3.6-Fix_Boot_Issue_On_Kvm-aCOSwt_P6.patch, metadata.xml: + Version bump. Bug #440116. Thanks to Eric F. GARIOUD + <eric-f.garioud@wanadoo.fr> +