WinXP crashes @ boot if GuestAdditions are installed (SMP/IO-APIC - AMD cpu)

[BenDJ]

on,

uname -ri

2.6.31.8-0.1-desktop x86_64

cat /proc/cpuinfo | grep "model name"

model name : AMD Phenom(tm) II X4 945 Processor

i've installed,

rpm -qa | grep -i virtualbox

VirtualBox-3.1-3.1.2_56127_openSUSE111-1.x86_64

with VT-x enabled.

i've installed WinXP Pro. the install goes fine, and the Guest (re)boots & functions normally.

if I install Guest Additions, the GA install completes ok. But, when I reboot the Guest, it FAILs, dropping me to the SafeMode etc selection screen. I can boot into SafeMode, but can't "Boot normally" at this point. It's not the Guest that's crashing, but rather the GuestOS ...

If I UNinstall GAs @ SafeMode desktop, thee reboot the machine, all's OK.

If I first DISable VT-x, then repeat -- installing GAs -- then, at reboot, the Guest boots fine.

Fully reproducible here.

[BenDJ]

just realized it's NOT the turning off of VT-x, but rather the reduction from 2 procs -> 1 proc that went along with it.

i.e.,

VT-x -> ON, # cpus == 1, boots OK with GAs VT-x -> ON, # cpus == 2, Windows fails to boot

[Technologov]

This means that the component must be "Guest SMP"

-Technologov

[Sander van Leeuwen]

Please check if 3.1.4 beta 1 solves the problem: ​http://forums.virtualbox.org/viewtopic.php?f=15&t=27300

[BenDJ]

installing,

rpm -Uvh VirtualBox-3.1-3.1.4_BETA1_57050_openSUSE112-1.x86_64.rpm

get a version # mismatch,

rpm -qa | grep -i VirtualBox

VirtualBox-3.1-3.1.4_BETA1_57050_openSUSE111-1.x86_64

openSUSE 111 vs 112 ...

but, launching VB with VT-d "ON", and any # cpus 1 -4, the GuestOS WinXP_Pro -- with GuestAdditions installed -- now launches OK.

one thing I _do_ notice is that time from launch to GuestOS Desktop seems significantly longer than b4. nothing quantitative, atm, though.

[BenDJ]

but, with the BETA, i can no longer get Display/Window resizing to work. If I zoom the container window to maximized - 1920 x 923, atm) the actual GuestOS display no longer changes with it.

Switching BACK to the release, non-bate version -- as well as switching back to 1-CPU -- cures the problem.

[Frank Mehnert]

bendj, that version mismatch is not an issue, we are just maintaining the same package for openSUSE 11.1 and openSUSE 11.2.

Regarding your problem with resizing the guest window: Did resizing the guest window stopped at all or only if you switch to host maximum? How much video RAM did you set for this VM?

But finally at least your XP guest crashes are gone, correct?

[BenDJ]

Replying to frank:

bendj, that version mismatch is not an issue

Ok.

Regarding your problem with resizing the guest window: Did resizing the guest window stopped at all or only if you switch to host maximum?

Afaict, it stopped completely -- neither switching to host maximum, nor to any arbitrary size, seemed to work.

How much video RAM did you set for this VM?

The Maximum -- 128Mb.

But finally at least your XP guest crashes are gone, correct?

As a result of the #cpus > 1, yes - the crashes are gone. I may be imagining it, but GuestOS system responsiveness seemed slower with the Beta + 4 CPUS, compared to the Release Version + 1 CPUS.

[Frank Mehnert]

I will close this defect as the original problem seems solved. Please create new tickets for remaining problems with VBox 3.1.4.

[Steve Slaven]

I am having this same problem with 3.2.6, and only on a Phenom II. If I run an smp XP Pro SP3 guest on 3.2.6 with multiple cpus it boots fine, after installing the guest additions booting failed with a variety of different BSODs. Reducing the number of cpus to 1 allows booting again. However installing guest additions on an smp XP guest works fine on an AMD Athalon II X3. If I use an older version of the guest additions (pre 3.0, such as 2.2.4 or 1.6.6) then multi cpu on the Phenom II is fine, and I still get graphic resizing, seamless, mouse integration, etc, so this seems like it might be related to the guest additions smp fixes for amd chipsets maybe? I am reopening this ticket because I also tried 3.1.8 with the same results, so I believe this problem existed before but seems to be CPU specific so maybe was missed.

[Steve Slaven]

Same problem in virtualbox 3.2.8 using guest additions 3.2.8 on Phenom II, endless bsod loops at boot. Using pre-3.0 guest additions, e.g. 2.2.4 then smp is fine and the guest runs normally using as many processors as I can set.

[Steve Slaven]

This is a log of the bsod reboot loop using guest additions 3.2.8 on phenom II

[Steve Slaven]

This is a log of the same machine using same version of virtualbox but guest additions 2.2.4 allowing smp xp pro to run

[Sander van Leeuwen]

Does it make any difference if you disable nested paging?

[Steve Slaven]

No difference disabling nested paging in 3.2.8, downloaded 3.2.10 and tried again with the same results. Installing the guest additions using an smp enabled xp guest and it goes in to a bsod loop. Disabling nested paging in 3.2.10 also made no difference. This smp guest functions fine on other 64 bit cpus (core i5, amd athalon II x3, etc) Here is my cpu info for the problematic cpu:

model name : AMD Phenom(tm) II X6 1055T Processor

flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx mmxext fxsr_opt pdpe1gb rdtscp lm 3dnowext 3dnow constant_tsc rep_good pni monitor cx16 popcnt lahf_lm cmp_legacy svm extapic cr8_legacy abm sse4a misalignsse 3dnowprefetch osvw ibs skinit wdt

[Sander van Leeuwen]

What kind of blue screen is it? A screenshot might help.

[Steve Slaven]

I took 4 bsod screen shots and I'm going to attach them, they occur during the initial bootup before you get a desktop or login prompt immediately after rebooting from installing the guest additions. So it's install guest additions, reboot, watch the windows logo thing, bsod. Now if I reboot I get the prompt to start normally or safe mode, etc, selecting start normally will then show the windows logo briefly then bsod. At this point it just loops like that, each startup will ask about normal or safe or last good startup, and anything except last good startup will bsod.

[Steve Slaven]

bsod after first reboot after installing guest additions

[Steve Slaven]

second reboot after selecting "normal startup"

[Steve Slaven]

third bsod after selecting "normal startup"

[Steve Slaven]

fourth bsod after selecting "normal startup"

[Sander van Leeuwen]

Are you able to extract a minidump from the guest? Only the BSOD from the 1st and 3rd one are interesting. (0x0000008E bug checks)

You should be able to work around the problem by executing:

VBoxManage setextradata "VM name" "VBoxInternal/HWVirtExt/TPRPatchingEnabled" 0

[Steve Slaven]

It doesn't look like it's generating minidumps for the bsod's at bootup. I can force a bsod during a normal bootup using ctrl+scroll lock+scroll lock and get a minidump, then reboot, install the guest additions, reboot and get the blue screen again but then it doesn't generate a new minidump, even though I know they are configured and should work since I was able to force a minidump before installing the guest additions.

However the vboxmanage command does allow the system to boot normally, even with the latest guest additions installed, however it seems to run slower.

[john mac]

I have the same problem with 3.2.10 with XPSP3 under Linux on a x64 AMD; except that it doesn't matter how many CPUs are configured or whether or not VT-x is enabled. Same BSOD; same values.

$ cat /proc/cpuinfo | grep "model name"
model name : AMD Athlon(tm) 7850 Dual-Core Processor

$ uname -a
Linux host 2.6.32-24-generic #43-Ubuntu SMP Thu Sep 16 14:58:24 UTC 2010 x86_64 GNU/Linux

$ dpkg -l | grep virtualbox
ii  virtualbox-3.2     3.2.10-66523~Ubuntu~lucid    Oracle VM VirtualBox
rc  virtualbox-ose     3.1.6-dfsg-2ubuntu2          x86 virtualization solution - base binaries
rc  virtualbox-ose-qt  3.1.6-dfsg-2ubuntu2          x86 virtualization solution - Qt based user 

[Steve Slaven]

I tried again with the latest version (3.2.12) of the guest additions with the same results, installing guest additions and enabling more than 1 cpu results in bsod loop

[john mac]

I replaced my AMD Athlon II X2 CPU with an AMD Phenon II X6 CPU and the problem disappeared.

This is using the exact same hardware, software, and virtual images in both cases (expect the CPU).

There's something about the architecture on the Athlon CPU that VirtualBox implements incorrectly.

[Frank Mehnert]

Please reopen if still relevant with VBox 4.1.6.