[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"



Andrew Lyon wrote:

How do you introduce the delay? I'd like to try xen-unstable with
viridian=1 but I want to be absolutely sure that the issue is fixed so
if you have a patch for qemu-dm I'd appreciate it if you could post
it, the BSOD 101 issue causes serious problems for me because I am
using pci passthrough, I've found  if the domain using passthru
crashes, I issue "xm pci-detach", or "xm destroy" a domain that is
using a passthru pci device, will cause the entire machine to lockup ,
I've tried Xen 3.2.1/2/3 and 3.3.0/1 with various different kernels
but it always happens, so if my Vista or 2008 domains that are using
pci passthru usb2 cards BSOD at all it takes down the entire box,
usually resulting in data loss :(

Perhaps that problem might be fixed in unstable too, I will try it
tomorrow but it would really help to have a way to trigger a bsod 101.

I use the attached change to qemu. In cpu_handle_ioreq, if send_vcpu != 0,, it checks for the presence of a file called /tmp/qemu_delay. If it exists, it opens it and reads a delay value from it. Then it sleeps for that duration (in microseconds).

What I do is boot windows 2008, then open an editor to edit /tmp/qemu_delay. From the editor, I write various values to it. An easy way to trigger the issue is to write a large value (say, 8000000 for 8 seconds), wait a little while, and then write 0 to make the domain come back to life. It should bluescreen at this point. Sometimes the procedure has to be repeated a few times.

If you want to do "echo $value > /tmp/qemu_delay", you'll need to change the hack to re-open the file each time, probably.

- Frank
*** target-i386-dm/helper2.c.orig       Wed Jan  7 12:37:30 2009
--- target-i386-dm/helper2.c    Mon Dec 22 16:09:55 2008
***************
*** 47,52 ****
--- 47,54 ----
  #include <limits.h>
  #include <fcntl.h>
  
+ #include <unistd.h>
+ 
  #include <xenctrl.h>
  #include <xen/hvm/ioreq.h>
  
***************
*** 570,575 ****
--- 572,580 ----
      extern int shutdown_requested;
      CPUState *env = opaque;
      ioreq_t *req = cpu_get_ioreq();
+     static int fd = -1;
+     char buf[16];
+     useconds_t udelay;
  
      handle_buffered_io(env);
      if (req) {
***************
*** 606,611 ****
--- 611,625 ----
        }
  
          req->state = STATE_IORESP_READY;
+       if (send_vcpu != 0) {
+           if (fd == -1)
+               fd = open("/tmp/qemu_delay", O_RDONLY);
+           if (fd >= 0) {
+               pread(fd, buf, sizeof buf, 0L);
+               udelay = atoi(buf);
+               usleep(udelay);
+           }
+       }
          xc_evtchn_notify(xce_handle, ioreq_local_port[send_vcpu]);
      }
  }
_______________________________________________
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®.