Hi all, I understand that currently the standard hibernation mechanisms of operating systems cannot be used with Xen. With Xen, it is possible to use xm save/xm restore which is in its effect quite similar, but I am a bit scared that this might cause problems within the affected OS because after restore, the system clock will show a great leap. This is also true with "native" hibernation, but the OS is aware of this and can react accordingly when it is woken up again. Does anybody know if this really can cause trouble or if - in the PV case - the kernel is patched so to that similar actions are taken as when the OS resumes from hibernation? Recently, I saw a patch on the xen-devel mailinglist that adds S3 support.>From my understanding, the potential problems that I have described abovewould be circumvented if the DomU is put into S3 state and then saved using xm save? Thanks for your patience, dla _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> I understand that currently the standard hibernation mechanisms of > operating systems cannot be used with Xen. > With Xen, it is possible to use xm save/xm restore which is in its effect > quite similar, but I am a bit scared that this might cause problems within > the affected OS because after restore, the system clock will show a great > leap. This is also true with "native" hibernation, but the OS is aware of > this and can react accordingly when it is woken up again. > Does anybody know if this really can cause trouble or if - in the PV case - > the kernel is patched so to that similar actions are taken as when the OS > resumes from hibernation?Paravirtualised guests are aware of the suspend / resume and are required to actively co-operate with it; therefore they ought to work fine. This doesn''t stop userspace apps getting a bit confused by the jump in time, but this is no more the case for Xen than any OS suspend / resume mechanism.> Recently, I saw a patch on the xen-devel mailinglist that adds S3 support. > > >From my understanding, the potential problems that I have described above > > would be circumvented if the DomU is put into S3 state and then saved using > xm save?Suspend / resume for HVM guests is in 3.1 but I''m not sure if or how it handles things like time jumps from the guest''s PoV. S3 isn''t currently used as part of this; I''m not entirely sure what''s going on in the power management area, but I do know there are general efforts to allow host low power states to work. Cheers, Mark -- Dave: Just a question. What use is a unicyle with no seat? And no pedals! Mark: To answer a question with a question: What use is a skateboard? Dave: Skateboards have wheels. Mark: My wheel has a wheel! _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> I am a bit scared that this might cause problems within the affected > OS because after restore, the system clock will show a great leap.IMHO, it would be really weird for an OS, apps or anything to seriously freak out if the clock leaped. I mean, does your computer crash a daylight savings? What about when you just set the clock by hand? If that breaks your OS or apps, you''ve got bigger problems. -Dylan _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
> > I am a bit scared that this might cause problems within the affected > > OS because after restore, the system clock will show a great leap. > > IMHO, it would be really weird for an OS, apps or anything to > seriously freak out if the clock leaped. I mean, does your computer > crash a daylight savings? What about when you just set the clock by > hand? If that breaks your OS or apps, you''ve got bigger problems.Well, if apps break they''re arguably badly written. Not that we can probably rule that out ;-) But I wouldn''t expect big problems with this; as Dylan quite rightly points out there are lots of circumstances when the clock can change unexpectedly. The case of suspend / resume is a little different, in that if you suspend an OS to disk and then resume it later without telling it it /could/ conceivably confuse the OS, which might not expect time to suddenly change underneath it. For PV, this is fine. For HVM... well I assume it works somehow ;-) I''m just not sure how it''s currently implemented or if it could maybe be improved by making the guest OS more aware of the suspend process. Cheers, Mark -- Dave: Just a question. What use is a unicyle with no seat? And no pedals! Mark: To answer a question with a question: What use is a skateboard? Dave: Skateboards have wheels. Mark: My wheel has a wheel! _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Here''s another thing. AFAIK, your running OS probably ignores the hardware clock. The only time your system accesses the hw clock is during boot up to set the system clock and during shutdown to set the HW clock to match up with the system clock. Your suspended guest should need NTP or something to correct the system clock when it wakes up. Disclaimer: I''m basing this off of experience setting up NTP servers 5 years ago, with zero exposure to how Windows handles things. I think it''s the same though.> > > I am a bit scared that this might cause problems within the affected > > > OS because after restore, the system clock will show a great leap. > > > > IMHO, it would be really weird for an OS, apps or anything to > > seriously freak out if the clock leaped. I mean, does your computer > > crash a daylight savings? What about when you just set the clock by > > hand? If that breaks your OS or apps, you''ve got bigger problems. > > Well, if apps break they''re arguably badly written. Not that we can probably > rule that out ;-) But I wouldn''t expect big problems with this; as Dylan > quite rightly points out there are lots of circumstances when the clock can > change unexpectedly. > > The case of suspend / resume is a little different, in that if you suspend an > OS to disk and then resume it later without telling it it /could/ conceivably > confuse the OS, which might not expect time to suddenly change underneath it. > For PV, this is fine. For HVM... well I assume it works somehow ;-) I''m > just not sure how it''s currently implemented or if it could maybe be improved > by making the guest OS more aware of the suspend process. > > Cheers, > Mark > > > -- > Dave: Just a question. What use is a unicyle with no seat? And no pedals! > Mark: To answer a question with a question: What use is a skateboard? > Dave: Skateboards have wheels. > Mark: My wheel has a wheel!_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
On Thu, Jul 12, 2007 at 10:58:06AM -0700, Dylan Martin wrote:>Disclaimer: I''m basing this off of experience setting up NTP servers 5 >years ago, with zero exposure to how Windows handles things. I think >it''s the same though.Alas, it doesn’t. Windows always accesses the hw clock. If you change the timezone, the hw clock will be changed. If correct the clock via NTP, the hw clock will be changed immediatley. Windows doesn’t have a system clock. And yes, I hate it. Shade and sweet water! Stephan -- | Stephan Seitz E-Mail: Nur-Ab-Sal@gmx.de | | PGP Public Keys: http://fsing.rootsland.net/~stse/pgp.html | _______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users
Hello all, thanks for the clarification. You saved me from reading the authoritative documentation (aka sourcecode) :-) By default, PV guests share the host''s time that cannot be changed from the DomU. The application layer is no problem as this can be put into the design requirements. My concern was the one that Mark pointed out and I am glad to hear that PV domains are prepared to handle this. However, advances in time should be less a problem than stepping backwards, as in that case there might be files that have been created or accessed in the "future". btw: If you have a look at tools like ntpdate, they can be configured to not correct the time immediately but over multiple iterations that only cause moderate time changes. Regards, dla On 7/12/07, Dylan Martin <dmartin@sccd.ctc.edu> wrote:> > Here''s another thing. AFAIK, your running OS probably ignores the > hardware clock. The only time your system accesses the hw clock is > during boot up to set the system clock and during shutdown to set the > HW clock to match up with the system clock. Your suspended guest > should need NTP or something to correct the system clock when it wakes > up. > > Disclaimer: I''m basing this off of experience setting up NTP servers 5 > years ago, with zero exposure to how Windows handles things. I think > it''s the same though. > > > > > I am a bit scared that this might cause problems within the affected > > > > OS because after restore, the system clock will show a great leap. > > > > > > IMHO, it would be really weird for an OS, apps or anything to > > > seriously freak out if the clock leaped. I mean, does your computer > > > crash a daylight savings? What about when you just set the clock by > > > hand? If that breaks your OS or apps, you''ve got bigger problems. > > > > Well, if apps break they''re arguably badly written. Not that we can > probably > > rule that out ;-) But I wouldn''t expect big problems with this; as > Dylan > > quite rightly points out there are lots of circumstances when the clock > can > > change unexpectedly. > > > > The case of suspend / resume is a little different, in that if you > suspend an > > OS to disk and then resume it later without telling it it /could/ > conceivably > > confuse the OS, which might not expect time to suddenly change > underneath it. > > For PV, this is fine. For HVM... well I assume it works somehow > ;-) I''m > > just not sure how it''s currently implemented or if it could maybe be > improved > > by making the guest OS more aware of the suspend process. > > > > Cheers, > > Mark > > > > > > -- > > Dave: Just a question. What use is a unicyle with no seat? And no > pedals! > > Mark: To answer a question with a question: What use is a skateboard? > > Dave: Skateboards have wheels. > > Mark: My wheel has a wheel! >_______________________________________________ Xen-users mailing list Xen-users@lists.xensource.com http://lists.xensource.com/xen-users