xm save -c problem

xm save -c problem

[Liu.yi]

hi,all
    I have a winxpsp2 vm with gplpv driver installed. The problem is that
when I excute "xm save -c winxpsp2 save.bin", the command is excuted
successfully but the vm hang after issuing a disk operaton.
    After installing gplpv debug driver, from the qemu log I found vbd
resume failed, after a while windows issues disk reset command, then vm
hang.
    xm debug-keys shows as follows before "xm save -c":
(XEN)        1 [0/1]: s=3 n=0 d=0 p=72 x=1
(XEN)        2 [0/0]: s=3 n=0 d=0 p=71 x=0
(XEN)        3 [0/1]: s=2 n=0 d=0 x=0
(XEN)        4 [0/0]: s=6 n=0 x=0                   // pdo_event_channel
(XEN)        5 [0/0]: s=2 n=0 d=0 x=0             // suspend_evtchn
(XEN)        6 [0/0]: s=3 n=0 d=0 p=73 x=0     // vbd-768 event-channel
(XEN)        7 [0/0]: s=3 n=0 d=0 p=74 x=0     // vif-0 event-channel
    xm debug-keys shows as follows after "xm save -c":
(XEN)        1 [0/1]: s=3 n=0 d=0 p=72 x=1
(XEN)        2 [0/0]: s=3 n=0 d=0 p=71 x=0
(XEN)        3 [0/1]: s=2 n=0 d=0 x=0
(XEN)        4 [0/0]: s=6 n=0 x=0                   // pdo_event_channel
(XEN)        5 [0/0]: s=2 n=0 d=0 x=0             // suspend_evtchn
(XEN)        6 [0/0]: s=6 n=0 x=0                   // new pdo_event_channel
(XEN)        7 [0/0]: s=2 n=0 d=0 x=0             // new suspend_evtchn
(XEN)        8 [0/0]: s=3 n=0 d=0 p=? x=0       // new vbd-768 event-channel
                                                              // vif-0
resume don't start at all because resume hang at vbd

    blkback and blkif driver seem free their event-channel in
unbind_from_irqhandler when suspending, so when gplpv driver resumes it
starts allocing event-channel from 6. The strange things is that gplpv
driver allocs new pdo_event_channel and suspend_evtchn (6 and 7), the
previous ones retain active.
    I tried to unbind the old pdo_event_channel and suspend_evtchn, but
suspending vm hang. "xm save -c" wokrs if I reuse the old pdo_event_channel
and suspend_evtchn as follows:

  in evtchn.c:EvtChn_Init()
      KeInitializeEvent(&xpdd->pdo_suspend_event, SynchronizationEvent,
FALSE);
      if (xpdd->pdo_event_channel == 0) {
          KdPrint((__DRIVER_NAME "     create new pdo_event_channel\n"));
          xpdd->pdo_event_channel = EvtChn_AllocIpi(xpdd, 0);
      }
  in xenpci_fdo.c:XenPci_ConnectSuspendEvt()
      if (xpdd->suspend_evtchn == 0) {
          xpdd->suspend_evtchn = EvtChn_AllocUnbound(xpdd, 0);
          KdPrint((__DRIVER_NAME "     create new suspend event
channel\n"));
      } 

    The qemu log show vbd and vif resume successfully, vm runs fine.
    I'm not sure about these modifications to gplpv windows driver. I will
be grateful for any suggestions, thks.



--
View this message in context: http://xen.1045712.n5.nabble.com/xm-save-c-problem-tp5713082.html
Sent from the Xen - Dev mailing list archive at Nabble.com.

Re: xm save -c problem

[Liu.yi]

    Unfortunately xm restore failed because when vm restores(resume)
xpdd->pdo_event_channel and xpdd->suspend_evtchn aren't 0.
    EvtChn_Unbind and EvtChn_Close pdo_event_channel in EvtChn_Init when
xpdd->pdo_event_channel isn't 0;
    EvtChn_Unbind and EvtChn_Close suspend_evtchn in
XenPci_ConnectSuspendEvt when xpdd->suspend_evtchn isn't 0;
    After modifing pv driver as above, xm save -c and xm restore seems work
fine. Again, I'm not sure about these.



--
View this message in context: http://xen.1045712.n5.nabble.com/xm-save-c-problem-tp5713082p5713084.html
Sent from the Xen - Dev mailing list archive at Nabble.com.

Re: xm save -c problem with GPLPV drivers and winxp

[Pasi Kärkkäinen]

On Sun, Dec 16, 2012 at 01:29:51AM -0800, Liu.yi wrote:
> hi,all

Hello,

I added the important keywords to the subject..

-- Pasi

>     I have a winxpsp2 vm with gplpv driver installed. The problem is that
> when I excute "xm save -c winxpsp2 save.bin", the command is excuted
> successfully but the vm hang after issuing a disk operaton.
>     After installing gplpv debug driver, from the qemu log I found vbd
> resume failed, after a while windows issues disk reset command, then vm
> hang.
>     xm debug-keys shows as follows before "xm save -c":
> (XEN)        1 [0/1]: s=3 n=0 d=0 p=72 x=1
> (XEN)        2 [0/0]: s=3 n=0 d=0 p=71 x=0
> (XEN)        3 [0/1]: s=2 n=0 d=0 x=0
> (XEN)        4 [0/0]: s=6 n=0 x=0                   // pdo_event_channel
> (XEN)        5 [0/0]: s=2 n=0 d=0 x=0             // suspend_evtchn
> (XEN)        6 [0/0]: s=3 n=0 d=0 p=73 x=0     // vbd-768 event-channel
> (XEN)        7 [0/0]: s=3 n=0 d=0 p=74 x=0     // vif-0 event-channel
>     xm debug-keys shows as follows after "xm save -c":
> (XEN)        1 [0/1]: s=3 n=0 d=0 p=72 x=1
> (XEN)        2 [0/0]: s=3 n=0 d=0 p=71 x=0
> (XEN)        3 [0/1]: s=2 n=0 d=0 x=0
> (XEN)        4 [0/0]: s=6 n=0 x=0                   // pdo_event_channel
> (XEN)        5 [0/0]: s=2 n=0 d=0 x=0             // suspend_evtchn
> (XEN)        6 [0/0]: s=6 n=0 x=0                   // new pdo_event_channel
> (XEN)        7 [0/0]: s=2 n=0 d=0 x=0             // new suspend_evtchn
> (XEN)        8 [0/0]: s=3 n=0 d=0 p=? x=0       // new vbd-768 event-channel
>                                                               // vif-0
> resume don't start at all because resume hang at vbd
> 
>     blkback and blkif driver seem free their event-channel in
> unbind_from_irqhandler when suspending, so when gplpv driver resumes it
> starts allocing event-channel from 6. The strange things is that gplpv
> driver allocs new pdo_event_channel and suspend_evtchn (6 and 7), the
> previous ones retain active.
>     I tried to unbind the old pdo_event_channel and suspend_evtchn, but
> suspending vm hang. "xm save -c" wokrs if I reuse the old pdo_event_channel
> and suspend_evtchn as follows:
> 
>   in evtchn.c:EvtChn_Init()
>       KeInitializeEvent(&xpdd->pdo_suspend_event, SynchronizationEvent,
> FALSE);
>       if (xpdd->pdo_event_channel == 0) {
>           KdPrint((__DRIVER_NAME "     create new pdo_event_channel\n"));
>           xpdd->pdo_event_channel = EvtChn_AllocIpi(xpdd, 0);
>       }
>   in xenpci_fdo.c:XenPci_ConnectSuspendEvt()
>       if (xpdd->suspend_evtchn == 0) {
>           xpdd->suspend_evtchn = EvtChn_AllocUnbound(xpdd, 0);
>           KdPrint((__DRIVER_NAME "     create new suspend event
> channel\n"));
>       } 
> 
>     The qemu log show vbd and vif resume successfully, vm runs fine.
>     I'm not sure about these modifications to gplpv windows driver. I will
> be grateful for any suggestions, thks.
> 
> 
> 
> --
> View this message in context: http://xen.1045712.n5.nabble.com/xm-save-c-problem-tp5713082.html
> Sent from the Xen - Dev mailing list archive at Nabble.com.
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel

Re: xm save -c problem with GPLPV drivers and winxp

[Konrad Rzeszutek Wilk]

On Sun, Dec 16, 2012 at 02:46:02PM +0200, Pasi Kärkkäinen wrote:
> On Sun, Dec 16, 2012 at 01:29:51AM -0800, Liu.yi wrote:
> > hi,all
> 
> Hello,
> 
> I added the important keywords to the subject..
> 
> -- Pasi
> 
> >     I have a winxpsp2 vm with gplpv driver installed. The problem is that
> > when I excute "xm save -c winxpsp2 save.bin", the command is excuted
> > successfully but the vm hang after issuing a disk operaton.
> >     After installing gplpv debug driver, from the qemu log I found vbd
> > resume failed, after a while windows issues disk reset command, then vm
> > hang.
> >     xm debug-keys shows as follows before "xm save -c":
> > (XEN)        1 [0/1]: s=3 n=0 d=0 p=72 x=1
> > (XEN)        2 [0/0]: s=3 n=0 d=0 p=71 x=0
> > (XEN)        3 [0/1]: s=2 n=0 d=0 x=0
> > (XEN)        4 [0/0]: s=6 n=0 x=0                   // pdo_event_channel
> > (XEN)        5 [0/0]: s=2 n=0 d=0 x=0             // suspend_evtchn
> > (XEN)        6 [0/0]: s=3 n=0 d=0 p=73 x=0     // vbd-768 event-channel
> > (XEN)        7 [0/0]: s=3 n=0 d=0 p=74 x=0     // vif-0 event-channel
> >     xm debug-keys shows as follows after "xm save -c":
> > (XEN)        1 [0/1]: s=3 n=0 d=0 p=72 x=1
> > (XEN)        2 [0/0]: s=3 n=0 d=0 p=71 x=0
> > (XEN)        3 [0/1]: s=2 n=0 d=0 x=0
> > (XEN)        4 [0/0]: s=6 n=0 x=0                   // pdo_event_channel
> > (XEN)        5 [0/0]: s=2 n=0 d=0 x=0             // suspend_evtchn
> > (XEN)        6 [0/0]: s=6 n=0 x=0                   // new pdo_event_channel
> > (XEN)        7 [0/0]: s=2 n=0 d=0 x=0             // new suspend_evtchn
> > (XEN)        8 [0/0]: s=3 n=0 d=0 p=? x=0       // new vbd-768 event-channel
> >                                                               // vif-0
> > resume don't start at all because resume hang at vbd
> > 
> >     blkback and blkif driver seem free their event-channel in
> > unbind_from_irqhandler when suspending, so when gplpv driver resumes it
> > starts allocing event-channel from 6. The strange things is that gplpv
> > driver allocs new pdo_event_channel and suspend_evtchn (6 and 7), the
> > previous ones retain active.
> >     I tried to unbind the old pdo_event_channel and suspend_evtchn, but
> > suspending vm hang. "xm save -c" wokrs if I reuse the old pdo_event_channel
> > and suspend_evtchn as follows:
> > 
> >   in evtchn.c:EvtChn_Init()
> >       KeInitializeEvent(&xpdd->pdo_suspend_event, SynchronizationEvent,
> > FALSE);
> >       if (xpdd->pdo_event_channel == 0) {
> >           KdPrint((__DRIVER_NAME "     create new pdo_event_channel\n"));
> >           xpdd->pdo_event_channel = EvtChn_AllocIpi(xpdd, 0);
> >       }
> >   in xenpci_fdo.c:XenPci_ConnectSuspendEvt()
> >       if (xpdd->suspend_evtchn == 0) {
> >           xpdd->suspend_evtchn = EvtChn_AllocUnbound(xpdd, 0);
> >           KdPrint((__DRIVER_NAME "     create new suspend event
> > channel\n"));
> >       } 
> > 
> >     The qemu log show vbd and vif resume successfully, vm runs fine.
> >     I'm not sure about these modifications to gplpv windows driver. I will
> > be grateful for any suggestions, thks.
> > 
> > 
> > 

CC-ing the author of the driver (James).