Hi, all
Some comments about test:
Tested the four patch under following HVM image, based on cs17589
1. Fc6 32/32e guest with both txt/X windows mode, no-vtd.
2. Fc6 32/32e guest with both txt/X windows mode, vtd-enabled,
assign one network card to the guest.
3. Fc5 32e HVM guest with PV driver, both txt/x windows mode.
It works fine. Test including little stress test,
such as S3 several times.
Thanks& Regards,
Criping
[PATCH 1/4] xen interface for HVM S3
- add two sub hypercalls under HVMOP hypercall:
HVMOP_s3_suspend/HVM_s3_resume
- HVMOP_s3_suspend will reset all HVM vcpus; meanwhile, power off all
the HVM virtual device in hypervisor, e.g, pic, apic, RTC and HPET
- HVMOP_s3_resume will set BSP eip to 0xfff0, cs base to 0xf0000, and
also set other registers/msr to the correct value/attributes in real
mode environment, so HVM will start from rombios post entry directly;
meanwhile, power on the HVM virtual device.
Signed-off-by: Tian Kevin <kevin.tian@intel.com>
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
Patches 2 and 3 look okay. Patch 4 is not really useful as it is, as you pointed out. And that then raises the question of what *is* the usage scenario for this feature? This also raises questions about patch 1 -- as it is I think it will need some cleaning up, but if we only wanted to do S3 across guest save/restore then we may not need Xen changes at all. We simply would not save/restore HVM sate but instead have tools manufacture a correct initial HVM state to load during restore. Corresponding to a clean initialised system with BP about to enter the BIOS. -- Keir On 8/5/08 09:30, "Ke, Liping" <liping.ke@intel.com> wrote:> Hi, all > Some comments about test: > Tested the four patch under following HVM image, based on cs17589 > 1. Fc6 32/32e guest with both txt/X windows mode, no-vtd. > 2. Fc6 32/32e guest with both txt/X windows mode, vtd-enabled, > assign one network card to the guest. > 3. Fc5 32e HVM guest with PV driver, both txt/x windows mode. > It works fine. Test including little stress test, > such as S3 several times. > Thanks& Regards, > Criping_______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
Keir Fraser wrote:> Patches 2 and 3 look okay. Patch 4 is not really useful as it is, as > you pointed out. And that then raises the question of what *is* the > usage scenario for this feature? This also raises questions about > patch 1 -- as it is I think it will need some cleaning up, but if we > only wanted to do S3 across guest save/restore then we may not need > Xen changes at all. We simply would not save/restore HVM sate but > instead have tools manufacture a correct initial HVM state to load > during restore. Corresponding to a clean initialised system with BP > about to enter the BIOS. > > -- Keir >Yes, this is good question. For the usage scenario, I agree with what Ian Pratt has pointed out. The most important one is the PCI pass through. Think about if one HVM domain is directly assigned with physical device, meanwhile, user what to do host S3. In this case, Xen must notify HVM domain to do virtual S3, otherwise, HVM domain will be surprise to see the assigned device is not working. With the evolution of the PCI passthrouhg technology, e.g. SR-IOV, we anticipate this scenario will be more and more comon especially when Xen is used in client side. Another usage scenario may be extended save/restore. To be honest, we did not anticipate this usage when design the virtual S3 feature, but surely we are happy if this feature can be used in this way, which means this feature is more useful than we expected :) For patch 1~3, they are basic infrastructure, virtual S3 can not be achieved without these 3 patches. And please feel free to have cleanup on patch1. For the patch4, it is mostly a user interface to call underlying resume infrastructure which is provided by patch 1~3, we are open on how this user interface could be. Patch4 just demonstrate how the underlying infrastructure can be used. And as Ian Pratt has said, probably we can also implemented "virtual lid" for this purpose in the next step. Best Regards Ke _______________________________________________ Xen-devel mailing list Xen-devel@lists.xensource.com http://lists.xensource.com/xen-devel
[PATCH 1/4] xen interface for HVM S3
- add two sub hypercalls under XEN_DOMCTL_sendtrigger hypercall:
XEN_DOMCTL_SENDTRIGGER_S3SUSPEND/XEN_DOMCTL_SENDTRIGGER_S3SUSPEND
- XEN_DOMCTL_SENDTRIGGER_S3SUSPEND will reset all HVM vcpus, pic, apic,
RTC and HPET. Set BSP eip to 0xfff0, cs base to 0xf0000, and
also set other registers/msr to the correct value/attributes in real
mode environment, so HVM will start from rombios post entry directly;
Then it pause the domain.
- XEN_DOMCTL_SENDTRIGGER_S3RESUME will unpause the domain
- add is_s3_suspend flag in domain for logging s3 suspend state
Signed-off-by: Tian Kevin <kevin.tian@intel.com>
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