Bug 643295

Description of problem:
After installing xenpv-win driver completely, select "reboot now ", the win-guest do "shutting down " and then it halt there with a gray screen.
After selecting uninstall and guest have to reboot. It do "shutting down" and then it halt there with a gray screen, same as rebooting after installing.


Version-Release number of selected component (if applicable):
xen:        3.0.3-117.el5
xenpv-win:  1.3.0-1.el5
kernel-xen: 2.6.18-226.el5
Host:       RHEL5.5-x86_64
Guest:      hvm-win2003-32


How reproducible:
100%

Steps to Reproduce:
1.create a win2003-32 hvm guest; 
2.install xenpv-win driver in the win-guest and reboot, as request.
3.boot a win2003-32 hvm guest with xenpv-win driver correctly
4.uninstall the xenpv-win driver in the win-guest and reboot, as request.


Actual results:
When rebooting, the win-guest do "shutting down" and then it halt there with a gray screen.


Expected results:
Win-guest reboot normally.


Additional info:

Created attachment 453672 [details]
xm dmesg for win2003-32 installation reboot

Created attachment 453674 [details]
xm dmesg for win2003-32 uninstallation reboot

Created attachment 453676 [details]
win2003-32 guest's config file

 no xend message in /var/log/xen/xend.log

This happens with multiprocessor guests.

In one occurrence, CPU2 shutdown properly, while CPU1 and CPU3 were executing something here:

rip: f75a9ca2 
rsp: f78bed34
rax: a2307548   rbx: f7737ee0   rcx: f7737ee0   rdx: 0000028b
rsi: f7737ec0   rdi: 89d71188   rbp: f78bed50
 r8: 00000000    r9: 00000000   r10: 00000000   r11: 00000000
r12: 00000000   r13: 00000000   r14: 00000000   r15: 00000000
 cs: 00000008    ds: 00000000    fs: 00000000    gs: 00000000

where edx:eax seems to be the TSC and everything is stuck at that IP.

In the meanwhile, CPU0 was stuck waiting for hal!HalpProcessorsNotHalted to become zero at 80a5ebae in the dump below:

hal!HalpResetThisProcessor:
80a5eb3e mov     edi,edi
80a5eb40 push    ebp
80a5eb41 mov     ebp,esp
80a5eb43 push    ecx
80a5eb44 mov     eax,dword ptr fs:[0000001Ch]
80a5eb4a mov     eax,dword ptr [eax+20h]
80a5eb4d movsx   eax,byte ptr [eax+10h]
80a5eb51 mov dword ptr ds:[0FFFE0080h],0FFh
80a5eb5b push    ebx
80a5eb5c mov dword ptr ds:[0FFFE0320h],3001Fh
80a5eb66 mov dword ptr ds:[0FFFE0350h],1001Fh
80a5eb70 xor     ebx,ebx
80a5eb72 push    esi
80a5eb73 mov dword ptr ds:[0FFFE0360h],184FFh
80a5eb7d cmp     dword ptr [hal!HalpMpInfoTable (80a6ca40)],ebx
80a5eb83 push    edi
80a5eb84 je      hal!HalpResetThisProcessor+0x52 (80a5eb90)
80a5eb86 mov dword ptr ds:[0FFFE0370h],100E3h
80a5eb90 cmp     eax,ebx
80a5eb92 jne     hal!HalpResetThisProcessor+0x166 (80a5eca4)
80a5eb98 mov     eax,offset hal!HalpProcessorsNotHalted (80a65ce4)
80a5eb9d mov     ecx,eax
80a5eb9f or      edx,0FFFFFFFFh
80a5eba2 lock xadd dword ptr [ecx],edx
80a5eba6 xor     ecx,ecx
80a5eba8 mov     edx,eax
80a5ebaa lock xadd dword ptr [edx],ecx
80a5ebae test    ecx,ecx
80a5ebb0 jne     hal!HalpResetThisProcessor+0x68 (80a5eba6)

I don't think it's particularly related to xenpv-win, changing component to kernel-xen.

*** This bug has been marked as a duplicate of bug 703693 ***

*** This bug has been marked as a duplicate of bug 701608 ***