On 28/01/2015 00:10, Paul Gortmaker wrote:
> Hi Greg,
>
> Someone working on a 3.10.x based -rt kernel traced a latency increase
> back to the backport of 56cc2406d68c0f0950 ("KVM: nVMX: fix "acknowledge
> interrupt on exit" when APICv is in use").
FWIW, that's fairly impossible. That commit does not execute any code
unless you are tinkering with nested virtualization.
Greg, I don't think it has to be reverted.
When they asked me for
> assistance, I noticed that the commit had:
>
> Fixes: 77b0f5d67ff2781f36831cba79674c3e97bd7acf
>
> ...but 77b0f5 only appeared in v3.16-rc4~21^2^2~32. Checking the stable
> queue, it seems that it was added to 3.10.54, 3.14.18 and 3.16.2 -- but if
> the Fixes tag is to be believed, then only 3.16.x should have this applied.
That said, when I add "Fixes" headers, the idea is that the patch
does/should not need backporting to stable unless the referenced commit
exists in that tree.
Paolo
> I also checked to see if stable [you] backported 77b0f5d67ff to any
> versions, thus expanding its applicability, but it/you had not.
>
> Obviously a kernel based on -rt won't behave the same as vanilla stable
> but regardless I figured I'd mention the discrepancy I found above since
> it might cause issues for other people on the stable kernels as well.
>
> Thanks,
> Paul.
> --
>
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html