Re: [PATCH 0/4] Kernel side patches for in-kernel APIC

[Avi Kivity <avi-atKUWr5tajBWk0Htik3J/w@public.gmane.org>]

Gregory Haskins wrote:
> The following is my kernel side patch series for adding in-kernel APIC logic.
> By default, the code enters "level-0" mode and should be compatible with
> existing userspace.  I have a patch series for userspace which enables
> "level-1" mode which I will forward after this one.
>
> I have incorporated most of the feedback I have received to date.  There were
> a few things that I had initially agreed to do that you may find missing from
> the changes.  I found a few places where my original decisions made more sense
> to me than what I agreed to change, so I left them pending further
> discussion.  E.g. "kvm_irqpin_t" was going to change to "kvm_cpuirq_t" but I
> decided against it for reasons I can discuss if anyone is so inclined.
>   

Please do (it doesn't really matter; just curious).  Also please provide 
in the future a changelog relative to previous patches so that it is 
easier to review.  It's fairly difficult to keep track of think with a 
long release cycle.


> level-0 has been tested with both 32 bit windows and 64-bit linux *before* I
> moved to git-HEAD.  They both worked without any discernable differences in
> behavior.
>
> I then bumped up to git-HEAD and adjusted all my patches to get ready for
> submission.  Unfortunately I seem to have run into a (known/unknown)
> regression(*) in the KVM codebase with that update where things arent working
> quite right.  What I did confirm was that the system behaves the same both
> with and without my patches for both level-0 and level-1 behavior.  It is now
> at a point where testers could start to look at my patches and provide
> bug/performance feedback in addition to the code review comments.  All are
> welcome.  
>   

Yes, we did have some regressions; I think I fixed all of them but of 
course I can't tell for sure.  Please try with latest HEAD, and if not 
let me know.

I'll review once I manage to get kvm-22 out; probably Monday or 
Tuesday.  Sorry about the delay.

Incidentally the latest changes should make the performance advantages 
of in-kernel lapic more noticable (or maybe, noticable) as the 
difference between guest->kernel switch times and guest->userspace 
switch times is increasing.



-- 
error compiling committee.c: too many arguments to function


-------------------------------------------------------------------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and take
control of your XML. No limits. Just data. Click to get it now.
http://sourceforge.net/powerbar/db2/