From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jeroen Groenewegen van der Weyden Subject: Re: Re: [Xen-users] Nested Virtualisation question Date: Sat, 09 Apr 2011 15:32:56 +0200 Message-ID: <4DA06008.2060700@grosc.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Keir Fraser Cc: "Christoph.Egger@amd.com" , "xen-devel@lists.xensource.com" , "Dong, Eddie" , Tim Deegan List-Id: xen-devel@lists.xenproject.org I'm now testing on a intel q9450, which does not seem to have this. I will move my testing to a intel e5645 which seems to have this capability. Question then is, when will the intel support be implemented? is there commitment from some developer? is there a roadmap planned. in short what is the story here. mfg, Jeroen Op 9-4-2011 13:15, Keir Fraser schreef: > On 09/04/2011 11:18, "Jeroen Groenewegen van der Weyden" > wrote: > >> The HAP is default set to one, I check this also with a dry run. >> Please confirm all this support just amd, because I only have Intel >> overhere. > Well, quite apart from being AMD only right now (Intel patches have been > posted in the past, but they're not in tree yet, and I very much doubt old > patches would directly apply to our current development tree), I'm also > pretty sure that your machine doesn't support Intel EPT -- if the processor > can't support HAP then we silently discard that configuration request when > creating a domain. That's why you get error 22 (EINVAL). But even if you did > have Intel EPT, because Intel's nested virt patches aren't in tree yet you'd > just get error 95 (EOPNOTSUPP) instead! > > By the way you can tell if your system supports EPT by grepping in 'xm > dmesg' output (i.e., Xen's boot log) for a line stating "Extended Page > Tables (EPT)". > > -- Keir > >> mfg, >> Jeroen >> >> Op 9-4-2011 8:22, Keir Fraser schreef: >>> On 09/04/2011 06:07, "Jeroen Groenewegen van der Weyden" >>> wrote: >>> >>>> The result with c/s23197 >>>> >>>> Regular domain now start, en stop normaly, (xl seems to have a problem >>>> though, after domain closes a domain null remains in de xl list/ xmdoe >>>> not have this.) >>>> for nestedVM I added the "nestedhvm=1" in the config file of the domain. >>>> but the domain does not start because of "invalid argument" >>> NestedHVM currently requires you to be using an AMD processor with hardware >>> nested paging (known as HAP by Xen). Most likely you need to add hap=1 to >>> your config file (and hope your processor supports it). >>> >>> -- Keir >>> >>>> +++++++++ beging console >>>> hydra:/etc/xen/vm # xm create groscNestedVM >>>> Using config file "./groscNestedVM". >>>> Error: (22, 'Invalid argument') >>>> +++++++++ end console >>>> >>>> +++++++++ begin xend.log >>>> hydra:/etc/xen/vm # tail /var/log/xen/xend.log >>>> dominfo = XendDomainInfo.create(config) >>>> File "/usr/lib64/python2.7/site-packages/xen/xend/XendDomainInfo.py", >>>> line 106, in create >>>> vm.start() >>>> File "/usr/lib64/python2.7/site-packages/xen/xend/XendDomainInfo.py", >>>> line 473, in start >>>> XendTask.log_progress(0, 30, self._constructDomain) >>>> File "/usr/lib64/python2.7/site-packages/xen/xend/XendTask.py", line >>>> 209, in log_progress >>>> retval = func(*args, **kwds) >>>> File "/usr/lib64/python2.7/site-packages/xen/xend/XendDomainInfo.py", >>>> line 2598, in _constructDomain >>>> xc.hvm_set_param(self.domid, HVM_PARAM_NESTEDHVM, long(nestedhvm)) >>>> Error: (22, 'Invalid argument') >>>> +++++++++++ end xend.log >>>> >>>> >>>> Op 7-4-2011 12:29, Tim Deegan schreef: >>>>> At 08:14 +0100 on 07 Apr (1302164093), Keir Fraser wrote: >>>>>> On 07/04/2011 08:09, "Keir Fraser" wrote: >>>>>>> Probably the same as this, happening often in the automated tests: >>>>>> Likely reason: nestedhvm_vcpu_initialise() being called unconditionally in >>>>>> hvm_vcpu_initialise() even though nestedhvm has not been configured for >>>>>> this >>>>>> domain. This then fails because there is no nestedhvm handler for Intel >>>>>> yet. >>>>> This bug should be fixed by 23175:e5a750d1bf9b, arriving at a staging >>>>> tree near you at any moment. >>>>> >>>>>> And beyond that it's most likely an issue of a broken failure path, or a >>>>>> problem in the HAP code with cleaning up a half-created guest. >>>>> This one I'll look into now. >>>>> >>>>> Tim. >>>>> >>>>> >>> > >