Xen devel - Jul 2013 - [BUG REPORT] Xen 4.3 xl failed to save domu with -c param

Hi all!

I think there is a bug in liibxl_dom.c
for this function -> libxl__domain_resume_device_model

976 int libxl__domain_resume_device_model(libxl__gc *gc, uint32_t domid)
977 {
978 
979     switch (libxl__device_model_version_running(gc, domid)) {
980     case LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN_TRADITIONAL: {
981         libxl__qemu_traditional_cmd(gc, domid, "continue");
982         libxl__wait_for_device_model(gc, domid, "running", NULL,
NULL, NULL);
983         break;
984     }
985     case LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN:
986         if (libxl__qmp_resume(gc, domid))
987             return ERROR_FAIL;
        break;
988     default:
989         return ERROR_INVAL;
990     }
991 
992     return 0;
993 }

we should add "break" between line 987 & 988

Another problem:

The guest hung after  cmd "xl save -c domid /fpath" with PV driver
if we remove/disable the pv driver, the function works well

Test ENV:
host: xen 4.3 + upstream qemu dm
dom0: Oracle UEK kernel
guest: win2k3 SP2 & centos 6.4

Additional
There is no valuable debug info in xl dmesg, /var/log/xen*

Any Suggestion?? 






rwxybh

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel

On Mon, Jul 29, 2013 at 10:57 AM, rwxybh <rwxybh@126.com>
wrote:
> Hi all!
>
> I think there is a bug in liibxl_dom.c
> for this function -> libxl__domain_resume_device_model
>
> 976 int libxl__domain_resume_device_model(libxl__gc *gc, uint32_t domid)
> 977 {
> 978
> 979     switch (libxl__device_model_version_running(gc, domid)) {
> 980     case LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN_TRADITIONAL: {
> 981         libxl__qemu_traditional_cmd(gc, domid, "continue");
> 982         libxl__wait_for_device_model(gc, domid, "running",
NULL, NULL,
> NULL);
> 983         break;
> 984     }
> 985     case LIBXL_DEVICE_MODEL_VERSION_QEMU_XEN:
> 986         if (libxl__qmp_resume(gc, domid))
> 987             return ERROR_FAIL;
>         break;
> 988     default:
> 989         return ERROR_INVAL;
> 990     }
> 991
> 992     return 0;
> 993 }
>
> we should add "break" between line 987 & 988
>
> Another problem:
>
> The guest hung after  cmd "xl save -c domid /fpath" with PV
driver
> if we remove/disable the pv driver, the function works well
Which PV driver are you using?

I think the PV drivers cooperate with the save / restore by
disconnecting and re-connecting; it''s possible that the
''-c'' parameter
isn''t sending one of the necessary signals.

 -George