Bug 521794

Hi Gleb
The time drift after Live migration still exist. it is not so serious though
Is it acceptable in the patch ?

Intel Host
Window 2k8 standard (64bits)
+--------+--------------------+--------------------+--------------------+
|        |    105.el5_4.2     |    105.el5_4.3 (1) |    105.el5_4.3 (2) |
+--------+--------------------+--------------------+--------------------+
|  boot  |   -                |  -                 |   -                |
+--------+--------------------+--------------------+--------------------+
| 10 mins|   1.242648         |   0.073379         |   -                |
+--------+--------------------+--------------------+--------------------+
| 20 mins|   2.631598         |   0.142490         |   -                |
+--------+--------------------+--------------------+--------------------+
| 30 mins|   6.893323         |   0.160291         |   -                |
+--------+--------------------+--------------------+--------------------+
| reboot |   -0.363271        |   -0.448923        |     1.060342       |
+--------+--------------------+--------------------+--------------------+
| 10 mins|1.947018(after5min) |   -0.357876        |     1.015323       |
+--------+--------------------+--------------------+--------------------+
| 20 mins|   -                |   -0.463631        |     0.986450       |
+--------+--------------------+--------------------+--------------------+
| 30 mins|   -                |   -0.373732        |     1.103563       |
+--------+--------------------+--------------------+--------------------+
|migrate |   19.764600        |   1.487834         |     9.449639       |
+--------+--------------------+--------------------+--------------------+
| 10 mins|   -                |   1.527333         |     9.441567       |
+--------+--------------------+--------------------+--------------------+
| 20 mins|   -                |   1.615365         |     9.414171       |
+--------+--------------------+--------------------+--------------------+
| 30 mins|   -                |   1.638770         |     9.560496       |
+--------+--------------------+--------------------+--------------------+ 


AMD host
Window XP
+--------+--------------------+--------------------+
|        | 105.el5_4.2        |  105.el5_4.3(1)    |
+--------+--------------------+--------------------+
|  boot  |  0.540955          |   -                |
+--------+--------------------+--------------------+
| 10 mins|2.077061 (after5min)|   0.073379         |
+--------+--------------------+--------------------+
| 20 mins|  -                 |   0.142490         |
+--------+--------------------+--------------------+
| 30 mins|  -                 |   0.160291         |
+--------+--------------------+--------------------+
| reboot |   5.455597         |   0.826407         |
+--------+--------------------+--------------------+
| 10 mins|   12.740380        |   0.765594         |
+--------+--------------------+--------------------+
| 20 mins|   -                |   0.873515         |
+--------+--------------------+--------------------+
| 30 mins|   -                |   0.791465         |
+--------+--------------------+--------------------+
|migrate |    23.421626       |   10.750137        | 
+--------+--------------------+--------------------+
| 10 mins|    27.744350       |   10.654840        |
+--------+--------------------+--------------------+
| 20 mins|   -                |   10.639801        |
+--------+--------------------+--------------------+
| 30 mins|   -                |   10.949871        |
+--------+--------------------+--------------------+

This patch has nothing to do with migration. Migration is different issue that  has (or should have) different BZ. Don't test this patch with migration.

(In reply to comment #6)
> Hi Gleb
> The time drift after Live migration still exist. it is not so serious though
> Is it acceptable in the patch ?
> 
> Intel Host
> Window 2k8 standard (64bits)

[root@dhcp-66-70-46 ~]# grep "model name"  /proc/cpuinfo
model name      : Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz
model name      : Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz
model name      : Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz
model name      : Intel(R) Core(TM)2 Quad CPU    Q9400  @ 2.66GHz

[root@dhcp-66-70-46 ~]# free -m
             total       used       free     shared    buffers     cached
Mem:          7727       7681         46          0         71        239
-/+ buffers/cache:       7369        357
Swap:         7959       2472       5487

Host is loaded.
for cpu in `seq 0 3`; do taskset -c $cpu yes >/dev/null & ; done

CLI : /usr/libexec/qemu-kvm -smp 2 -m 4G -drive file=/data/images/win2008-64-virtio.raw,if=ide -net nic,macaddr=00:15:00:62:AF:02,model=virtio -net tap -vnc :12 -name t77 -monitor stdio -rtc-td-hack -startdate now -usbdevice tablet


> 
> AMD host
> Window XP
[root@dhcp-66-70-48 ~]# grep "model name"  /proc/cpuinfo 
model name      : AMD Phenom(tm) 9600B Quad-Core Processor
model name      : AMD Phenom(tm) 9600B Quad-Core Processor
model name      : AMD Phenom(tm) 9600B Quad-Core Processor
model name      : AMD Phenom(tm) 9600B Quad-Core Processor

[root@dhcp-66-70-48 ~]# free  -m
             total       used       free     shared    buffers     cached
Mem:          7590       4993       2596          0         59        658
-/+ buffers/cache:       4275       3315
Swap:         7891         41       7850

Host is loaded.
for cpu in `seq 0 3`; do taskset -c $cpu yes >/dev/null & ; done

CLI :/usr/libexec/qemu-kvm -smp 2 -m 4G -cpu qemu64,+sse2 -startdate now -drive file=/mnt/winXP-32-virtio-back.raw,media=disk,if=ide -name host2guest -usbdevice tablet -net nic,vlan=0,macaddr=00:55:18:02:B2:fe,model=virtio -net tap,vlan=0,script=/etc/qemu-ifup -vnc :2 -monitor stdio -notify all -rtc-td-hack


=========
steps:
A : 
1 start vm
2 synchronize guest's time with host  ( ntpdate.exe -b $HOST )
3 have the guest ran 30mins. query the offset. ( ntpdate.exe -q $HOST )

B :
1 synchronize guest's time with host  ( ntpdate.exe -b $HOST )
2 restart guest.
3 have the guest ran 30mins. query the offset. ( ntpdate.exe -q $HOST )

C :
1 synchronize guest's time with host  ( ntpdate.exe -b $HOST )
2 implement migration ( src and dst are the same host )
3 have the guest ran 30mins. query the offset. ( ntpdate.exe -q $HOST )

(In reply to comment #7)
> This patch has nothing to do with migration. Migration is different issue that 
> has (or should have) different BZ. Don't test this patch with migration.  

New bug about migration is reported :
https://bugzilla.redhat.com/show_bug.cgi?id=523457

Created attachment 362296 [details]
521794 result

Test more guest on AMD and Intel Host .
Attached is the result .

and the more information can be found from :
http://focus.bne.redhat.com/~lihuang/521974/

compared with the result in result in comment#6.
the rtc-td-hack take effect in kvm-83-105.el5_4.4

Hi gleb. 

I noticed sometimes window guest's time  is 1.5~2 seconds later the host's after starting 
It is acceptable ?

Is this acceptable I don't know. PM should decide. But this is known problem. During boot windows keeps rtc interrupt masked, so it miss many interrupts since we don't redeliver interrupt that was issued while irq was masked by a guest.

Yes . have a bug for rhel guest.(523478)
but don't if we have a workaround for window guest .

so in this bz. we concern about if rtc-td-hack take effect or not . right ?

yes.(In reply to comment #15)
> Yes . have a bug for rhel guest.(523478)
> but don't if we have a workaround for window guest .
> 
> so in this bz. we concern about if rtc-td-hack take effect or not . right ?  

yes

Created attachment 362851 [details]
result in kvm-83-105.el5_4.7

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHSA-2009-1465.html