[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Xen-devel] Re: BSOD "A clock interrupt was not recevied onasecondary processor within the allocated time interval"




This is very similar to the patch we have on top of the original Novell stuff.

Steve Ofsthun was digging through the new viridian.c code at the end of the day yesterday to see if this bit was set. This fix below should do the same thing.



-----Original Message-----
From: Steven Smith [mailto:steven.smith@xxxxxxxxxxxxx]
Sent: Tue 1/6/2009 5:39 AM
To: Keir Fraser
Cc: Ben Guthro; Steve Prochniak; Andrew Lyon; Xen-Devel (E-mail); xen-users@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-devel] Re: BSOD "A clock interrupt was not recevied onasecondary processor within the allocated time interval"

> Alternative Viridian interface support was checked in. When enabled, it
> ought to be sufficient to disable these bugchecks. Oviridian=1¹ needs to be
> specified in the domain config file.
Hmm...  In order for the Viridian stuff to actually solve this
problem, you need to set the relaxed-timers bit.  It doesn't look like
the xen-unstable implementation does so.  Something like this might
help:

diff -r f6b92526e916 xen/arch/x86/hvm/viridian.c
--- a/xen/arch/x86/hvm/viridian.c       Tue Jan 06 09:14:39 2009 +0000
+++ b/xen/arch/x86/hvm/viridian.c       Tue Jan 06 10:32:26 2009 +0000
@@ -37,6 +37,7 @@

 /* Viridian CPUID 4000004, Implementation Recommendations. */
 #define CPUID4A_MSR_BASED_APIC  (1 << 3)
+#define CPUID4A_RELAX_TIMER_INT_HANDLING (1 << 5)

 int cpuid_viridian_leaves(unsigned int leaf, unsigned int *eax,
                           unsigned int *ebx, unsigned int *ecx,
@@ -84,7 +85,7 @@
         if ( (d->arch.hvm_domain.viridian.guest_os_id.raw == 0) ||
              (d->arch.hvm_domain.viridian.guest_os_id.fields.os < 4) )
             break;
-        *eax = CPUID4A_MSR_BASED_APIC;
+        *eax = CPUID4A_MSR_BASED_APIC | CPUID4A_RELAX_TIMER_INT_HANDLING;
         *ebx = 2047; /* long spin count */
         break;
     }

But note that I haven't tested this *at all*.

Steven.


> On 05/01/2009 21:32, "Ben Guthro" <bguthro@xxxxxxxxxxxxxxx> wrote:
>
> > This was never accepted by upstream Xen, as originally submitted by Ky from
> > Novell:
> >
> > original submission:
> > http://lists.xensource.com/archives/html/xen-devel/2008-02/msg00630.html
> >
> >
> >
> > Steve Prochniak wrote on 01/05/2009 03:55 PM:
> >> 
> >> Andrew - you can run the "Novell Shim" which makes the hypervisor
> >> conform to the Microsoft spec enough that the guest OSs will turn off
> >> watchdog timeouts (0x101 BSoDs)...
> >>
> >> Steve
> >>
> >> -----Original Message-----
> >> From: Andrew Lyon [mailto:andrew.lyon@xxxxxxxxx]
> >> Sent: Monday, January 05, 2009 3:24 PM
> >> To: Steve Prochniak; Xen-Devel (E-mail); xen-users@xxxxxxxxxxxxxxxxxxx
> >> Subject: Re: [Xen-devel] Re: BSOD "A clock interrupt was not recevied
> >> onasecondary processor within the allocated time interval"
> >>
> >> On Mon, Jan 5, 2009 at 7:25 PM, Steve Prochniak
> >> <sprochniak@xxxxxxxxxxxxxxx> <mailto:sprochniak@xxxxxxxxxxxxxxx>  wrote:
> >>  
> >> 
> >>> 
> >>> All Microsoft 6.0 and beyond Operating Systems have a sense of
> >>> 'enlightment' which means they try to talk to the underlying
> >>>    
> >>> 
> >> 
> >> hypervisor
> >>  
> >> 
> >>> 
> >>> - and if there is one, they do certain things like turn off 101
> >>> bugchecks.  The only real way to get Vista and beyond to work
> >>>    
> >>> 
> >> 
> >> correctly
> >>  
> >> 
> >>> 
> >>> with SMP is to make Xen conform to Microsoft's hypervisor spec.
> >>>    
> >>> 
> >> 
> >>
> >> I vaguely recall there being a way to make xen "lie" to the OS about
> >> the type of hypervisor, it might only be a feature on the commercial
> >> citrix xenserver?
> >>
> >> So basically there is no way to run Windows Vista or 2008 on Xen
> >> without risk of a 101 BSOD under load?
> >>
> >> If that is true then its a severe problem and Xen is useless for newer
> >> MS OS's :(
> >>
> >> Has anybody else had this problem and found a solution?
> >>
> >> Andy
> >>
> >>
> >>  
> >> 
> >>> 
> >>> -----Original Message-----
> >>> From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx
> >>> [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of Andrew
> >>>    
> >>> 
> >> 
> >> Lyon
> >>  
> >> 
> >>> 
> >>> Sent: Friday, January 02, 2009 1:34 PM
> >>> To: xen-users@xxxxxxxxxxxxxxxxxxx; Xen-Devel (E-mail)
> >>> Subject: [Xen-devel] Re: BSOD "A clock interrupt was not recevied
> >>> onasecondary processor within the allocated time interval"
> >>>
> >>> On Mon, Dec 29, 2008 at 9:32 AM, Andrew Lyon <andrew.lyon@xxxxxxxxx>
> >>> <mailto:andrew.lyon@xxxxxxxxx>
> >>> wrote:
> >>>    
> >>> 
> >>>> 
> >>>> Hi,
> >>>>
> >>>> When dom0 is under heavy load any Vista or Windows 2008 HVM's that
> >>>>      
> >>>> 
> >>> 
> >> 
> >> are
> >>  
> >> 
> >>> 
> >>>> 
> >>>> running and have multiple cpu's assigned often  BSOD with code
> >>>> 0x00000101 "A clock interrupt was not recevied ona  secondary
> >>>> processor within the allocated time interval"
> >>>>
> >>>> It only happens if the load in dom0 is high enough to make the mouse
> >>>> pointer lagged, once the mouse fails to track in realtime the hvm's
> >>>> start to bsod within a few seconds.
> >>>>
> >>>> I have tried several versions of Xen including 3.2, 3.3 and 3.3.1
> >>>>      
> >>>> 
> >>> 
> >> 
> >> rc4,
> >>  
> >> 
> >>> 
> >>>> 
> >>>> and various kernels including the 2.6.18 xensource and 2.6.27-xen.hg.
> >>>>
> >>>> Here is a example config from one of my vista hvms, is there a
> >>>> timer/rtc setting that I need to add to avoid this problem?
> >>>>
> >>>> import os, re
> >>>> arch = os.uname()[4]
> >>>> if re.search('64', arch):
> >>>>    arch_libdir = 'lib64'
> >>>> else:
> >>>>    arch_libdir = 'lib'
> >>>> kernel = "/usr/lib/xen/boot/hvmloader"
> >>>> builder='hvm'
> >>>> memory = 2048
> >>>> name = "Vistax86"
> >>>> uuid = "b7bd2f2f-169f-4789-8aee-eaa77c543c99"
> >>>> vcpus=8
> >>>> vif = [ 'mac=00:16:3e:7d:bc:b1' ]
> >>>> disk = [ 'phy:/dev/vg_raptor/lv_vistax86,hda,w', ',hdc:cdrom,r' ]
> >>>> device_model = '/usr/' + arch_libdir + '/xen/bin/qemu-dm'
> >>>> boot="d"
> >>>> sdl=0
> >>>> vnc=1
> >>>> vnclisten="127.0.0.1"
> >>>> vncdisplay=11
> >>>> vncpasswd='vv8176a'
> >>>> stdvga=0
> >>>> serial='pty'
> >>>> usbdevice='tablet'
> >>>>
> >>>>      
> >>>> 
> >>> 
> >> 
> >> cpuid=['1:edx=xxx1xxxxxxxxxxxxxxxxxxxxxxxxxxxx,ebx=xxxxxxxx00010000xxxxx
> >>  
> >> 
> >>> 
> >>> xxxxxxxxxxx','4,0:eax=001111xxxxxxxxxxxxxxxxxxxxxxxxxx']
> >>>    
> >>> 
> >>>> 
> >>>> keymap='en-gb'
> >>>>
> >>>> Andy
> >>>>
> >>>>      
> >>>> 
> >>> 
> >>> I think this is a known issue which has incorrectly been marked as
> >>> FIXED in bugzilla:
> >>>
> >>> http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1117
> >>>
> >>> there is a second bug report of the same problem, this time with more
> >>> details:
> >>>
> >>> http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=1065
> >>>
> >>> Is there a fix?
> >>>
> >>> Andy
> >>>
> >>> _______________________________________________
> >>> Xen-devel mailing list
> >>> Xen-devel@xxxxxxxxxxxxxxxxxxx
> >>> http://lists.xensource.com/xen-devel
> >>>
> >>>    
> >>> 
> >> 
> >>
> >> _______________________________________________
> >> Xen-devel mailing list
> >> Xen-devel@xxxxxxxxxxxxxxxxxxx
> >> http://lists.xensource.com/xen-devel
> >>  
> >
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@xxxxxxxxxxxxxxxxxxx
> > http://lists.xensource.com/xen-devel
>

> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel

 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.