Discuss VirtualBox 2.1.4 Release

[TerryE]

This topic is for discussions on the VirtualBox 2.1.4 Release as announced today. For discussions on 2.1.2 see Discuss VirtualBox 2.1.2 Release. However, users are recommended to upgrade to 2.1.4 at the earliest convenience

[shoot-me]

- Installation went with out any troubles, on ubuntu 8.10.

- visual response performance of my xp sp2 guest seems to be better.

- re-installation of guest xp addons installed without problems.

- The screen hogging when of the xp guest when all the windows are minimized still occurs when in seemless mode.

- Haven't tried the clonezilla or any other cloning experiments yet, but I will later.

Keep up the good work !

[sej7278]

Wow seems to be a big regression with Solaris hosts and guests, and no mention of a fix for the gradual slowdown and eventual freezing of solaris 10 guests.

Updated fine on Fedora10/64-bit.

Guest additions installed ok in:

solaris 10u6
opensolaris 2008.11 (broke resizing)
win2003 sp1
win2003r2 sp2
winxp pro sp3
oracle (rhel) 5.3
centos 5.2
centos 4.7
debian 5.0

As the changelog says, seemless mode is pretty badly broken when using compiz. fullscreen and windowed mode work fine with guest resizing in windows (don't think fullscreen/seemless ever worked in solaris guests).

EDIT: just noticed that in OpenSolaris 2008.11, after install GA214, we no longer have resizable display, its stuck at 1024x768 even in fullscreen. going back to 2.1.2's additions makes it work again.

I'm convinced Sun are trying to make Solaris work badly with VirtualBox so we'll all move to Zones/Xen!

[warpcafe]

Heya,

I tried to download the 2.1.4 version for x86 Windows hosts, however it downloads only about 500+ K of an MSI file which (not surprisingly) turns out to be defective.
URL: http://download.virtualbox.org/virtualb ... in_x86.msi

I tried to download with a real browser (FFx3) as well as IE6 however the result was the same... anyone experienced the same issue?

TIA
Thomas

[warpcafe]

Hi again,

in reply to my own posting - I have managed to download it from Sun at
http://dlc.sun.com/virtualbox/2.1.4/Vir ... in_x86.msi
...also the GuestAdditions are available from there.

...in case someone experienced the same problems.

Cheers, thanks for your good work,
Thomas

[mark_orion]

I upgraded from 2.0.6 to 2.1.4 on Ubuntu 8.04 hardy heron. I am using the virtualbox repository:

Installation/Upgrade
- Upgrade process removed 2.0.6 and installed 2.1.4, but did not compile new kernel module. This had to be done manually via sudo /etc/init.d/vboxdrv setup.
- Am I right that /etc/init.d/vboxnet is obsolete since 2.1.0? The upgrade did not remove that initscript or the link in /etc/rc2.d
- Guest Addition upgrade on two XP guests went well with one guest freezing during the required reboot.

Performance
- Increased performance / boot speed of Windows XP guests

3D Acceleration
- Comparative test with OGLBench showed impressive framerates.
- Resizing of XP guest window with OpenGL active works - Good Work!
Some 3D apps behave strange:
Blender (www.blender.org) starts with warnings:
OpenGL Warning: Packer wont pass SelectBuffer()
OpenGL Warning: Try using feedbackspu
While Blenders 3D modeling GUI and the rendering works, the gameengine always crashes.
Windows built-in OpenGL screensavers resize screen to 640x480 when starting in preview mode.

Disappointment
The VDI compacting feature is still not available. I understand that the developers want to enable that feature for all supported virtual disk formats, but why are you not just leaving the VDI code in the build and work on the VMDK compacting instead of removing existing features.

Overall
Thanks for the good work and the improvements in performance and 3D. But please do not remove existing features for the sake of still non existing new ones.

[sej7278]

- Upgrade process removed 2.0.6 and installed 2.1.4, but did not compile new kernel module. This had to be done manually via sudo /etc/init.d/vboxdrv setup.

thats interesting as i noticed for the first time that "/etc/init.d/vboxdrv setup" was run automatically when i upgraded the rpm on fedora10.

- Increased performance / boot speed of Windows XP guests

i never saw a problem with speed, although i don't bother with xp as its kernel is not optimal for virtualising anyway, think that was a debian/ubuntu-only issue.

The VDI compacting feature is still not available.

yes it is a shame they've disabled a feature that previously worked just so that they can add something new, but it seems to be something that happens a lot with virtualbox, maybe the code isn't as modular as it should be.....

[manfredell]

This topic is for discussions on the VirtualBox 2.1.4 Release as announced today. For discussions on 2.1.2 see Discuss VirtualBox 2.1.2 Release. However, users are recommended to upgrade to 2.1.4 at the earliest convenience

On OSX: the bug of the GUI not remembering the last position of its window when starting (shifting to the bottom each time it is closed and restarted) is still there. Not nice.

[billq]

1st issue...

RHEL 5.2 server using XP guests with/without VT-x or PAE enabled experience 100% cpu even when XP is "idle". strace of and idle VBox PID is:

Code: Select all

futex(0x73b650c, FUTEX_WAIT, 16957, NULL) = 0
futex(0x73b6558, FUTEX_WAKE, 1)         = 0
futex(0x73b650c, FUTEX_WAIT, 16959, NULL) = 0
futex(0x73b6558, FUTEX_WAKE, 1)         = 0
futex(0x73b650c, FUTEX_WAIT, 16961, NULL) = 0
futex(0x73b6558, FUTEX_WAKE, 1)         = 0
futex(0x73b650c, FUTEX_WAIT, 16963, NULL) = 0
futex(0x73b6558, FUTEX_WAKE, 1)         = 0
futex(0x73b650c, FUTEX_WAIT, 16965, NULL) = 0
futex(0x73b6558, FUTEX_WAKE, 1)         = 0

2nd issue...

Code: Select all

Reboot of XP guest takes ~ 10 minutes.  strace of VBox PID spins on:
ioctl(38, FIONREAD, [0])                = 0
clock_gettime(CLOCK_MONOTONIC, {510821, 553203727}) = 0
clock_gettime(CLOCK_MONOTONIC, {510821, 553280727}) = 0
clock_gettime(CLOCK_MONOTONIC, {510821, 553355727}) = 0
ioctl(38, FIONREAD, [0])                = 0
poll([{fd=32, events=POLLIN}, {fd=34, events=POLLIN}, {fd=38, events=POLLIN}, {fd=42, events=POLLIN}], 4, 49) = 0 (Timeout)
clock_gettime(CLOCK_MONOTONIC, {510821, 603010727}) = 0
ioctl(38, FIONREAD, [0])                = 0
clock_gettime(CLOCK_MONOTONIC, {510821, 603211727}) = 0
clock_gettime(CLOCK_MONOTONIC, {510821, 603289727}) = 0
clock_gettime(CLOCK_MONOTONIC, {510821, 603363727}) = 0
ioctl(38, FIONREAD, [0])                = 0
poll([{fd=32, events=POLLIN}, {fd=34, events=POLLIN}, {fd=38, events=POLLIN}, {fd=42, events=POLLIN}], 4, 49) = 0 (Timeout)
clock_gettime(CLOCK_MONOTONIC, {510821, 653020727}) = 0
ioctl(38, FIONREAD, [0])                = 0

When the spinning stop, I see this:

Code: Select all

clock_gettime(CLOCK_MONOTONIC, {512775, 332469939}) = 0
clock_gettime(CLOCK_MONOTONIC, {512775, 332496939}) = 0
clock_gettime(CLOCK_MONOTONIC, {512775, 332520939}) = 0
ioctl(38, FIONREAD, [0])                = 0
poll([{fd=32, events=POLLIN}, {fd=34, events=POLLIN}, {fd=38, events=POLLIN}, {fd=42, events=POLLIN}], 4, 49) = 1 ([{fd=34, revents=POLLIN}])
clock_gettime(CLOCK_MONOTONIC, {512775, 339332939}) = 0
ioctl(38, FIONREAD, [0])                = 0
read(34, "\0", 16)                      = 1
read(34, 0x7fff2d1462c0, 16)            = -1 EAGAIN (Resource temporarily unavailable)
writev(45, [{"H\2\6\377\304\2\200\3\r\0\340\3\200\2f\0\0\0\0\0\0\30\5\0", 24}, {"\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 261120}], 2) = 261144

-Bill

[sej7278]

RHEL 5.2 server using XP guests with/without VT-x or PAE enabled experience 100% cpu even when XP is "idle".

don't think that's ever going to get fixed, its a longstanding issue, certainly not specific to 2.1.4

basically the acpi kernel of xp just isn't up to being virtualised - you get the same kind of problem with vmware.

you can replace it with a uniprocessor kernel and it helps decrease the load.

better to use win2003 as a guest.

[mcp]

Just installed 2.1.4 on Ubuntu hardy heron and it works great. With 2.1.2 my vm ran like a dog. They seemed to have fixed the slow down, at least for my programs. AutoCad 2009 works great.

Thanks for the great work,
MCP

[TerryE]

I've got two Ubuntu JeOS 32bit LAMP VMs running on my dual core AMD Ubuntu 64bit, at the moment with 2 SSH sessions onto the VMs and host as well as a VNC session onto the host display. One of the terminal sessions on the host display is running top and its ticking along at around 8% (out of 200%), so I am quite happy

[MartyS]

On OSX: the bug of the GUI not remembering the last position of its window when starting (shifting to the bottom each time it is closed and restarted) is still there. Not nice.

I have an open ticket on this for version 2.1.2 because while it's minor compared to some things, it is aggravating time and time again and it's the very first thing you see with the GUI.

Even if the window location won't be remembered between runs, if at least the window's starting position/size would better fit (move it to the very top of the screen if they have to) it would be an improvement.

/Marty

[MartyS]

I had been running the 2.1.2 release before I upgraded on my OSX host. After I ran the 2.1.4 upgrade utility I resumed a Window's XP SP3 client that previously hadn't exhibited any issues. The resume seemed okay, so I immediately went about upgrading the guest's additions. Quickly I started getting message popups telling me "delayed write failed" on several items of the quest additions. Then more of the same (error) for other applications running in the background.

I've tried installing the new guest additions now 5 times. I cannot get a complete run without at least some of these errors. I've shutdown and rebooted the guest several times too. Now I can't even get that far as the guest just "hangs" by the time I try and start the very first application. The guest boots just fine -- it's after my guest desktop appears that my issues come up.

If I could back off some of the settings, optimizations, etc. I would do so but I don't know where to turn, other than to re-install 2.1.2 but I don't know if that's advisable.

/Marty

[jrv]

FreeBSD guest still doesn't work (won't complete "make buildworld") due to the sigreturn / eflags bug. Mac OSX host.

That concerns me a little because I suspect that FreeBSD might be the only OS that notices and reports a problem with eflags being clobbered: I wonder if it's possible other guest OSs are getting clobbered eflags but don't report it.