[Patch 4/4]: Xend interface for HVM S3 - extend "xm resume <domid>" to be able to S3 resume HVM domain. - when user issue "xm resume" command for HVM domain, xend will use xc lib API to call HVMOP_s3_resume hypercall. Note: it may not appropriate use xm resume for HVM s3, since xm resume is originally designed for save/restor purpose. It will be fine that xend maintainer find another place to put this code. Signed-off-by: Yu Ke <ke.yu@intel.com> Signed-off-by: Ke Liping <liping.ke@intel.com> _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ke, Liping writes ("[Xen-devel] [Patch 4/4]: Xend interface for HVM S3"):> [Patch 4/4]: Xend interface for HVM S3 > - extend "xm resume <domid>" to be able to S3 resume HVM domain. > - when user issue "xm resume" command for HVM domain, xend will use > xc lib API to call HVMOP_s3_resume hypercall. > > Note: it may not appropriate use xm resume for HVM s3, since xm resume > is originally designed for save/restor purpose. It will be fine that > xend maintainer find another place to put this code.If you want to reuse `xm resume'' for this you need to think about the interactions with the existing `xm suspend'' and `xm resume''. What combination of states are permissible, what transitions exist, and how are they triggered ? What if the guest is in the middle of going into one kind of suspend when a request arrives to put it into a different kind of suspend ? (Users often make multiple similar requests if they think the first request isn''t working.) Ian. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Jackson wrote:> Ke, Liping writes ("[Xen-devel] [Patch 4/4]: Xend interface for HVM > S3"): >> [Patch 4/4]: Xend interface for HVM S3 >> - extend "xm resume <domid>" to be able to S3 resume HVM domain. >> - when user issue "xm resume" command for HVM domain, xend will >> use xc lib API to call HVMOP_s3_resume hypercall. >> >> Note: it may not appropriate use xm resume for HVM s3, since xm >> resume is originally designed for save/restor purpose. It will be >> fine that xend maintainer find another place to put this code. > > If you want to reuse `xm resume'' for this you need to think about the > interactions with the existing `xm suspend'' and `xm resume''. What > combination of states are permissible, what transitions exist, and how > are they triggered ?Since hvm suspend is initiated by HVM guest, so only xm resume entry is reused here.> > What if the guest is in the middle of going into one kind of suspend > when a request arrives to put it into a different kind of suspend ? > (Users often make multiple similar requests if they think the first > request isn''t working.)Currently it will do HVM_S3_Resume when we detect it is hvm domain in domain_resume temporarially. I think it is a temporary solution. I think there are only two combinations: if it is save, resume -> restore. Otherwise, resume->s3 wake up for HVM. So do you think which way is better, reuse xm resume or not?> Ian.Thanks a lot! Regards, Criping _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ke, Liping writes ("RE: [Xen-devel] [Patch 4/4]: Xend interface for HVM S3"):> Since hvm suspend is initiated by HVM guest, so only xm resume entry > is reused here.I see. That seems ... odd.> > What if the guest is in the middle of going into one kind of suspend > > when a request arrives to put it into a different kind of suspend ? > > (Users often make multiple similar requests if they think the first > > request isn''t working.) > > Currently it will do HVM_S3_Resume when we detect it is hvm domain > in domain_resume temporarially. I think it is a temporary solution. > I think there are only two combinations: if it is save, resume -> restore. > Otherwise, resume->s3 wake up for HVM.Can the guest be saved (with `xm suspend'') while it is suspending (into HVM S3) ? If so then you may have to resume it twice ...> So do you think which way is better, reuse xm resume or not?Sad though I am to say it, I think we need yet another kind of resumption/restoration/restarting/etc. Ian. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
On 8/5/08 11:13, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote:>> So do you think which way is better, reuse xm resume or not? > > Sad though I am to say it, I think we need yet another kind of > resumption/restoration/restarting/etc.This has always been the problem with the virtual S3 support. Actually usefully integrating it into the save/restore is the hard bit. -- Keir _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Stefan de Konink
2008-May-08 10:37 UTC
Re: [Xen-devel] [Patch 4/4]: Xend interface for HVM S3
On Thu, 8 May 2008, Keir Fraser wrote:> On 8/5/08 11:13, "Ian Jackson" <Ian.Jackson@eu.citrix.com> wrote: > > >> So do you think which way is better, reuse xm resume or not? > > > > Sad though I am to say it, I think we need yet another kind of > > resumption/restoration/restarting/etc. > > This has always been the problem with the virtual S3 support. Actually > usefully integrating it into the save/restore is the hard bit.Is there currently a practical use why virtual S3 is prefered over the save command? Is it more gracefully? (or is the actual usage the other way around, when a user initiates the S3) Stefan _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
> Is there currently a practical use why virtual S3 is prefered over the > save command? Is it more gracefully?They have quite different purposes. One use for guest S3 is in integrating it into a save/restore so that the guest isn''t so surprised when it wakes up and finds that time has suddenly warped forward. In practice, Linux/Windows seem to cope just fine without passing through S3. Doing so might be cleaner, but at the risk of tickling guest operating system S3 bugs Guest S3 is also useful in the context of PCI device passthrough, where its necessary to get the device into a safe state.> (or is the actual usage the other way around, when a user initiatesthe> S3)The guest must initiate S3, but you can typically prod it into doing so via the ''lid switch''. Ian> > > Stefan > > > _______________________________________________ > Xen-devel mailing list > Xen-devel@lists.xensource.com > http://lists.xensource.com/xen-devel_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Ian Pratt writes ("RE: [Xen-devel] [Patch 4/4]: Xend interface for HVM S3"):> Guest S3 is also useful in the context of PCI device passthrough, where > its necessary to get the device into a safe state.S3 doesn''t involve writing a memory image to disk; it''s `suspend to RAM'' - more like xm pause than xm save. S4 is suspend to disk, like xm save. xm pause is certainly a lot simpler than the full-on ACPI S3 support. I think the latter is, as Ian says, mainly useful when the guest has direct access to hardware devices. If you want to put the physical machine into S3 or S4 then you need cooperation from all the domains which own the hardware. Ian. _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
[PATCH 4/4]: Xend interface for HVM S3 - extend "xm trigger <domid> s3resume" to be able to S3 resume HVM domain. - when user issue "xm trigger" command for HVM domain, xend will use xc lib API to call XEN_DOMCTL_SENDTRIGGER_S3RESUME domctl hypercall. - add s3_suspended state for HVM domain Note: S3 suspend is initiated from HVM guest itself and resumed from xm trigger command. We also add a new HVM state: s3_suspended, so that save/restore now could work with a s3_suspended domain. Signed-off-by: Yu Ke <ke.yu@intel.com> Signed-off-by: Ke Liping <liping.ke@intel.com _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Seemingly Similar Threads
- [PATCH] ioemu-remote: ACPI S3 state wake up
- [pvops-dom0]Let PV ops guest could handle Machine Check trap
- FW: [patch 0/4]Enable CMCI (Corrected Machine Check Error Interrupt) for Intel CPUs
- [Doc] writeup for error handling usage in XEN
- [RFC][patch 0/7] Enable PCIE-AER support for XEN