[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH 0/6] x86/HVM: RTC/VPT adjustments

At 10:23 +0100 on 02 May (1367490201), George Dunlap wrote:
> On 02/05/13 07:58, Jan Beulich wrote:
> >>>>On 01.05.13 at 18:15, George Dunlap <george.dunlap@xxxxxxxxxxxxx> wrote:
> >>On 29/04/13 14:42, Jan Beulich wrote:
> >>>1: fix processing of RTC REG_B writes
> >>>2: slightly adjust RTC reset
> >>>3: adjust IRQ (de-)assertion
> >>>4: properly handle RTC periodic timer even when !RTC_PIE
> >>>5: fix legacy PIC check in pt_update_irq()
> >>>6: RTC code must be in line with WAET flags passed by hvmloader
> >>>
> >>>This fixes the Win2003 boot failure George reported. Roger, since
> >>>the first patch is slightly different from what you tested earlier,
> >>>could you re-test that patch alone and the full series against
> >>>FreeBSD?
> >>>
> >>>Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
> >>This series seems to fix the w2k3 issue -- but it looks like a series of
> >>"fixes and updates".  I thought we had decided to revert all the RTC
> >>changes?
> >I always said I'd prefer a partial revert over a full one, if at all
> >possible. Of course I'm not intending to enforce this in any way,
> >but I'm also not intending to myself revert good fixes (and drop
> >further ones, as presented in this series) without need. So my
> >proposed solution is this series of patches (which is a partial
> >revert in terms of functionality, but not any kind of revert in terms
> >of source code) - others can certainly propose other solutions.
> >This is even more so now that we know that the reason for the
> >observed regression weren't the RTC changes themselves, but
> >the expectation of non-spec-conforming emulation behavior by
> >the guest.
> OK -- well I'll leave it to you and Tim to judge; just let me remind you 
> of our primary goals at this point (in order of importance):
> 1. A bug-free 4.3 release
> 2. An awesome 4.3 release
> 3. An on-time 4.3 release
> And that for #1, in particular we're worried about bugs that that may 
> not be detected until after the release.

On those grounds, as I've said, I think we should back out the vpt/rtc
interaction changes.  They look likely enough to be correct, but we
though that before the last two bugs surfaced, so without some serious
testing I don't think they're ready.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.