Bug 1170937 - [Power KVM] Grub interface of guest response slow
Summary: [Power KVM] Grub interface of guest response slow
Keywords:
Status: CLOSED EOL
Alias: None
Product: RHEV for Power
Classification: Retired
Component: qemu-kvm-rhev
Version: unspecified
Hardware: ppc64
OS: Unspecified
unspecified
high
Target Milestone: rc
: ---
Assignee: David Gibson
QA Contact: Virtualization Bugs
URL:
Whiteboard:
Depends On: 1212256
Blocks: RHEV4.0PPC RHV4.1PPC
TreeView+ depends on / blocked
 
Reported: 2014-12-05 07:13 UTC by Gu Nini
Modified: 2016-07-25 14:18 UTC (History)
13 users (show)

Fixed In Version:
Clone Of:
: 1212256 (view as bug list)
Environment:
Last Closed: 2016-05-11 09:53:25 UTC
Embargoed:


Attachments (Terms of Use)
virsh_dumpxml-Bz1170937 (3.91 KB, text/plain)
2015-05-29 08:47 UTC, Gu Nini
no flags Details


Links
System ID Private Priority Status Summary Last Updated
IBM Linux Technology Center 121449 0 None None None Never

Description Gu Nini 2014-12-05 07:13:03 UTC
Description of problem:
When enter grub of the guest, the interface responses very slow

Version-Release number of selected component (if applicable):
Host kernel: 3.10.0-201.el7.ppc64
Guest kernel: 3.10.0-195.el7.ppc64/3.10.0-196.ael7a.ppc64le
qemu-kvm:
qemu-img-rhev-2.1.2-14.el7.ppc64
qemu-kvm-tools-rhev-2.1.2-14.el7.ppc64
qemu-kvm-common-rhev-2.1.2-14.el7.ppc64
qemu-kvm-rhev-2.1.2-14.el7.ppc64
qemu-kvm-rhev-debuginfo-2.1.2-14.el7.ppc64

How reproducible:
100%

Steps to Reproduce:
1. Define and start a guest with virsh command
2. After the guest boot up, reboot the guest to enter its grub interface
3. Try to type some words

Actual results:
In step2, it would take a while before the grub screen refresh the whole info
In step3, after type the words, the curser would move right and left in the line before it could find its right position

Expected results:
In step2, The grub srceen could refresh the whole info in a minute
In step3, after type any words, the curser could move correctly

Additional info:

Comment 2 Gu Nini 2015-02-06 03:00:22 UTC
The bug also exists on power system, the software versions are as follows:

Host kernel: 3.10.53-2020.1.pkvm2_1_1.49.ppc64
Qemu kvm:
qemu-2.0.0-2.1.pkvm2_1_1.20.40.ppc64
qemu-debuginfo-2.0.0-2.1.pkvm2_1_1.20.40.ppc64
qemu-common-2.0.0-2.1.pkvm2_1_1.20.40.ppc64
qemu-kvm-tools-2.0.0-2.1.pkvm2_1_1.20.40.ppc64
qemu-img-2.0.0-2.1.pkvm2_1_1.20.40.ppc64
qemu-kvm-2.0.0-2.1.pkvm2_1_1.20.40.ppc64
qemu-system-x86-2.0.0-2.1.pkvm2_1_1.20.40.ppc64
qemu-system-ppc-2.0.0-2.1.pkvm2_1_1.20.40.ppc64

Comment 3 IBM Bug Proxy 2015-02-06 13:10:34 UTC
------- Comment From fnovak.com 2015-02-06 13:07 EDT-------
reverse mirror of RHBZ 1170937 - [Power KVM] Grub interface of guest response slow

Comment 4 David Gibson 2015-02-16 22:31:37 UTC
What method are you using to access the VM's console?  Is it graphical or text?

Comment 5 Gu Nini 2015-02-28 02:12:49 UTC
(In reply to David Gibson from comment #4)
> What method are you using to access the VM's console?  Is it graphical or
> text?

That's text.

Comment 7 David Gibson 2015-04-16 02:54:39 UTC
Since this occurs on a RHEL host too, cloning to RHEL bug 1212256.

Comment 8 Thomas Huth 2015-05-27 06:06:38 UTC
A possible solution for this problem is described here:

https://lists.ozlabs.org/pipermail/linuxppc-dev/2015-May/129286.html

Comment 9 IBM Bug Proxy 2015-05-29 07:40:13 UTC
------- Comment From KURZGREG.com 2015-05-29 07:35 EDT-------
The final patchset was applied today to the upstream repository (URLs below):

https://github.com/aik/SLOF/commit/57ca8e04b426aed43d51eb67805abd5eb9c4449f

which reads "fbuffer: simplify address computations in fb8-toggle-cursor"

https://github.com/aik/SLOF/commit/4c7a70f0619534f5870d4043f1751259046db02f

which reads "fbuffer: introduce the invert-region helper"

https://github.com/aik/SLOF/commit/99c534ecc7a8566bd9ca6346915d9ac1bfacae1e

which reads "fbuffer: introduce the invert-region-x helper"

Cheers.

Comment 10 Thomas Huth 2015-05-29 08:24:23 UTC
Gu Nini, to be able to fully reproduce and verify the fix for this issue, I think I need the XML definition of the guest where you discovered this problem. Could you please attach it to this bug ticket (i.e. the output of "virsh dumpxml ...")?

Comment 11 Thomas Huth 2015-05-29 08:30:43 UTC
(In reply to Gu Nini from comment #5)
> (In reply to David Gibson from comment #4)
> > What method are you using to access the VM's console?  Is it graphical or
> > text?
> 
> That's text.

I assume that you you're using VNC to view the console of the guest? Or did you use "virsh console ..." to see it in your normal terminal window?

Comment 12 Gu Nini 2015-05-29 08:47:00 UTC
Created attachment 1031738 [details]
virsh_dumpxml-Bz1170937

Comment 13 Gu Nini 2015-05-29 08:50:08 UTC
(In reply to Thomas Huth from comment #11)
> (In reply to Gu Nini from comment #5)
> > (In reply to David Gibson from comment #4)
> > > What method are you using to access the VM's console?  Is it graphical or
> > > text?
> > 
> > That's text.
> 
> I assume that you you're using VNC to view the console of the guest? Or did
> you use "virsh console ..." to see it in your normal terminal window?

Yes, I have been using the VNC connection.
For the xml definition needed in comment 10, please refer to the attachment in comment 12.

Comment 14 Thomas Huth 2015-05-29 09:12:48 UTC
Ok, thanks for the information ... Now I am pretty sure that Greg Kurz's patches are the right fix for this problem (since these are about a fix in the firmware graphics card driver - which would not apply if you had used "virsh console ..." instead)

Comment 15 IBM Bug Proxy 2015-05-29 10:01:34 UTC
------- Comment From KURZGREG.com 2015-05-29 09:54 EDT-------
(In reply to comment #12)
> (In reply to Thomas Huth from comment #11)
> > (In reply to Gu Nini from comment #5)
> > > (In reply to David Gibson from comment #4)
> > > > What method are you using to access the VM's console?  Is it graphical or
> > > > text?
> > >
> > > That's text.
> > I assume that you you're using VNC to view the console of the guest? Or did
> > you use "virsh console ..." to see it in your normal terminal window?
> Yes, I have been using the VNC connection.
> For the xml definition needed in comment 10, please refer to the attachment
> in comment 12.

FWIW, a much simpler reproducer is:

qemu-system-ppc64 -enable-kvm -vnc :66 -cdrom /any/distro/install/iso

Comment 16 David Gibson 2015-06-01 00:25:39 UTC
IBM,

We're now in the process of fixing this on the RHEL side by applying Greg Kurz' patches.  You should do the same if you're not already.

For reference, the commits you need from Alexey's SLOF tree are:

57ca8e0 fbuffer: simplify address computations in fb8-toggle-cursor
4c7a70f fbuffer: introduce the invert-region helper
99c534e fbuffer: introduce the invert-region-x helper

Comment 17 Thomas Huth 2016-05-04 16:02:42 UTC
I think this has been fixed in the latest version of pkvm, at least SLOF-0.1.git20150601-1.1.pkvm2_1_1.12 seems to contain the "invert-region-x" fix. So is it OK to close this ticket now?

Comment 18 Thomas Huth 2016-05-10 11:25:35 UTC
Gu Nini, could you please check whether the latest version is working better now?

Comment 19 Gu Nini 2016-05-11 09:31:23 UTC
(In reply to Thomas Huth from comment #18)
> Gu Nini, could you please check whether the latest version is working better
> now?

Thomas,

This bug had been originally reported on ppc64 host; while it was copied to ppc64le host with bz 1212256, which was verified/closed after bisecting another 2 bugs bz 1237052(verified/closed) and bz 1248448(new). 

Thus, should we still check it on ppc64 host?

Comment 20 Thomas Huth 2016-05-11 09:53:25 UTC
Since BZ 1212256 has already been verified/closed, and ppc64le is now our main architecture for RHEV, I think it should be OK to close this bug directly. So I'm closing it as "EOL" now - if anybody disagrees, please feel free to open this ticket again (e.g. in case we need more detailed testing).


Note You need to log in before you can comment on or make changes to this bug.