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

RE: [Xen-devel] when timer go back in dom0 save and restore or migrate, PV domain hung



This is not a s/r or lm specific issue. For example, system time can be changed even when pv guest is running. Your patch only hacks restore point once, and wc_sec can still be changed later when system time is changed on-the-fly again.
 
IIRC, pv guest can catch up wall clock change in timer interrupt, and time_resume will sync internal processed system time with new system time after restored. But I'm not sure whether it's enough. Actually the more interesting is the uptime difference. For example, timer with expiration calculated on previous system time may wait nearly infinite if uptime among two boxes vary a lot. But I think such issue should have been considered already, e.g. some user tool assistance. I think Keir can comment better here.
 
BTW, do you happen to know what exactly dom0 hangs on? In some busy loop to catch up time, or long delay to some critical timer expiration?
 
Thanks,
Kevin


From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of James Song
Sent: Tuesday, November 25, 2008 4:02 PM
To: xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: [Xen-devel] when timer go back in dom0 save and restore or migrate, PV domain hung

Hi,
   I find PV domin hung, When we take those steps
         1, save PV domain
         2, change system time of PV domain back
         3, restore a PV domain
        or
         1, migrate a PV domain from Machine A to Machine B
         2, the system time of Machine B is slower than Machine A.
   the problem is wc_sec will be change when system-time chanaged in dom0 or restore in a slower-system-time machine, but when restoring, xen don't restore the wc_sec of share_info from xenstore and use native one. So guest os will hang.
this patch will work for this issue.

 Thanks
 -- Song Wei

diff -r a5ed0dbc829f tools/libxc/xc_domain_restore.c
--- a/tools/libxc/xc_domain_restore.c    Tue Nov 18 14:34:14 2008 +0800
+++ b/tools/libxc/xc_domain_restore.c    Fri Nov 21 17:34:15 2008 +0800
@@ -328,6 +328,16 @@
 
     /* For info only */
     nr_pfns = 0;
+    //jsong@xxxxxxxxxx, james song
+    memset(&domctl, 0, sizeof(domctl));
+    domctl.domain = dom;
+    domctl.cmd    = XEN_DOMCTL_restoredomain;
+    frc = do_domctl(xc_handle, &domctl);
+    if ( frc != 0 )
+    {
+            ERROR("Unable to set flag of restore.");
+            goto out;
+    }
 
     if ( read_exact(io_fd, &p2m_size, sizeof(unsigned long)) )
     {
@@ -1120,6 +1130,8 @@
 
     /* restore saved vcpu_info and arch specific info */
     MEMCPY_FIELD(new_shared_info, old_shared_info, vcpu_info);
+    MEMCPY_FIELD(new_shared_info, old_shared_info, wc_nsec);
+    MEMCPY_FIELD(new_shared_info, old_shared_info, wc_sec);
     MEMCPY_FIELD(new_shared_info, old_shared_info, arch);
 
     /* clear any pending events and the selector */
diff -r a5ed0dbc829f xen/arch/x86/time.c
--- a/xen/arch/x86/time.c    Tue Nov 18 14:34:14 2008 +0800
+++ b/xen/arch/x86/time.c    Fri Nov 21 17:34:15 2008 +0800
@@ -689,7 +689,6 @@
     wmb();
     (*version)++;
 }
-
 void update_vcpu_system_time(struct vcpu *v)
 {
     struct cpu_time       *t;
@@ -703,7 +702,6 @@
 
     if ( u->tsc_timestamp == t->local_tsc_stamp )
         return;
-
     version_update_begin(&u->version);
 
     u->tsc_timestamp     = t->local_tsc_stamp;
@@ -713,14 +711,19 @@
 
     version_update_end(&u->version);
 }
-
 void update_domain_wallclock_time(struct domain *d)
 {
     spin_lock(&wc_lock);
+    if(d->after_restore )
+    {
+        d->after_restore = 0;
+        goto out; //jsong@xxxxxxxxxx
+    }
     version_update_begin(&shared_info(d, wc_version));
     shared_info(d, wc_sec)  = wc_sec + d->time_offset_seconds;
     shared_info(d, wc_nsec) = wc_nsec;
     version_update_end(&shared_info(d, wc_version));
+out:
     spin_unlock(&wc_lock);
 }
 
@@ -751,7 +754,6 @@
     u64 x;
     u32 y, _wc_sec, _wc_nsec;
     struct domain *d;
-
     x = (secs * 1000000000ULL) + (u64)nsecs - system_time_base;
     y = do_div(x, 1000000000);
 
@@ -1050,7 +1052,6 @@
 struct tm wallclock_time(void)
 {
     uint64_t seconds;
-
     if ( !wc_sec )
         return (struct tm) { 0 };
 
diff -r a5ed0dbc829f xen/common/domctl.c
--- a/xen/common/domctl.c    Tue Nov 18 14:34:14 2008 +0800
+++ b/xen/common/domctl.c    Fri Nov 21 17:34:15 2008 +0800
@@ -24,7 +24,6 @@
 #include <asm/current.h>
 #include <public/domctl.h>
 #include <xsm/xsm.h>
-
 extern long arch_do_domctl(
     struct xen_domctl *op, XEN_GUEST_HANDLE(xen_domctl_t) u_domctl);
 
@@ -315,6 +314,16 @@
         ret = 0;
     }
     break;
+   case XEN_DOMCTL_restoredomain:
+   {
+        struct domain *d;
+        if ( (d = rcu_lock_domain_by_id(op->domain)) == NULL )
+            break;
+       
+        d->after_restore = 1;
+        rcu_unlock_domain(d);
+        break;
+   }
 
     case XEN_DOMCTL_createdomain:
     {
diff -r a5ed0dbc829f xen/include/public/domctl.h
--- a/xen/include/public/domctl.h    Tue Nov 18 14:34:14 2008 +0800
+++ b/xen/include/public/domctl.h    Fri Nov 21 17:34:15 2008 +0800
@@ -61,6 +61,7 @@
 #define XEN_DOMCTL_destroydomain      2
 #define XEN_DOMCTL_pausedomain        3
 #define XEN_DOMCTL_unpausedomain      4
+#define XEN_DOMCTL_restoredomain      51
 #define XEN_DOMCTL_resumedomain      27
 
 #define XEN_DOMCTL_getdomaininfo      5
diff -r a5ed0dbc829f xen/include/xen/sched.h
--- a/xen/include/xen/sched.h    Tue Nov 18 14:34:14 2008 +0800
+++ b/xen/include/xen/sched.h    Fri Nov 21 17:34:15 2008 +0800
@@ -231,6 +231,7 @@
      * cause a deadlock. Acquirers don't spin waiting; they preempt.
      */
     spinlock_t hypercall_deadlock_mutex;
+    int after_restore; //jsong@xxxxxxxxxx
 };
 
 struct domain_setup_info
---------------------------------------------------------------------------------------------
 Thanks
--Song wei

_______________________________________________
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®.