Virtualbox-4.3.0 - multiple hard crashes
-
- Posts: 3
- Joined: 14. Mar 2011, 15:02
- Primary OS: Ubuntu other
- VBox Version: PUEL
- Guest OSses: Windows 7, Windows XP
- Location: Scotland
Re: Virtualbox-4.3.0 - multiple hard crashes
Hi guys.
I noticed this in my system logs the other day. Seems to explain why the 4.3 series is having problems particularly with NAT in the guests.
NAT[1743]: segfault at 9cd43826 ip a547d956 sp 9b41cfc8 error 4 in VBoxDD.so[a535f000+270000]
Doesn't happen all the time, but NAT doesn't work either, the guest is not getting an IP address from the virtual DHCP server.
I wound back to 4.2.18 for the time being, as neither XP nor Mint are working properly under both 4.3.0 and 4.3.2.
I noticed this in my system logs the other day. Seems to explain why the 4.3 series is having problems particularly with NAT in the guests.
NAT[1743]: segfault at 9cd43826 ip a547d956 sp 9b41cfc8 error 4 in VBoxDD.so[a535f000+270000]
Doesn't happen all the time, but NAT doesn't work either, the guest is not getting an IP address from the virtual DHCP server.
I wound back to 4.2.18 for the time being, as neither XP nor Mint are working properly under both 4.3.0 and 4.3.2.
Re: Virtualbox-4.3.0 - multiple hard crashes
Sorry, but this isn't even remotely related to anything we discuss here.gordboy wrote:NAT[1743]: segfault at 9cd43826 ip a547d956 sp 9b41cfc8 error 4 in VBoxDD.so[a535f000+270000]
Re: Virtualbox-4.3.0 - multiple hard crashes
New kernel log with frame pointers. Not sure if it is of much help, the first trace is empty. I have also included the VM log and settings from VBoxManage showvminfo --details.saintdev wrote:I have recompiled with frame pointers and will attempt to get a backtrace when I'm prepared to lockup my system again.Code: Select all
$ uname -a Linux khopesh 3.10.17-ck.0 #2 SMP PREEMPT Tue Oct 22 16:26:22 MDT 2013 x86_64 Intel(R) Core(TM)2 Quad CPU Q9300 @ 2.50GHz GenuineIntel GNU/Linux
Edit: I am also willing to do more testing and/or supply the virtual machine. This VM can consistently lock up my host in a minute or two by just doing an update.
- Attachments
-
- khopesh-dmesg.log
- Kernel log
- (86.99 KiB) Downloaded 19 times
-
- vminfo.txt
- VM settings
- (3.44 KiB) Downloaded 16 times
-
- Arch Router Clone-2013-11-11-13-04-19.log
- VM log
- (57.7 KiB) Downloaded 15 times
Re: Virtualbox-4.3.0 - multiple hard crashes
An update of what?saintdev wrote:This VM can consistently lock up my host in a minute or two by just doing an update.
Re: Virtualbox-4.3.0 - multiple hard crashes
When I update the Arch guest.skoehler wrote:An update of what?saintdev wrote:This VM can consistently lock up my host in a minute or two by just doing an update.
All I have to do is run pacman -Syu and the guest is not even able to finish updating before causing a kernel panic on the host.
Re: Virtualbox-4.3.0 - multiple hard crashes
So it's a linux guest? There is a tool called "stress". It allows you to generate plenty of CPU and I/O load. Can you reproduce the crash with that?saintdev wrote:When I update the Arch guest.
All I have to do is run pacman -Syu and the guest is not even able to finish updating before causing a kernel panic on the host.
-
- Oracle Corporation
- Posts: 793
- Joined: 7. Jan 2008, 16:17
Re: Virtualbox-4.3.0 - multiple hard crashes
Code: Select all
[131856.147729] huh, entered softirq 1 TIMER ffffffff8104a350 preempt_count 00000200, exited with 00000101?
[131856.147735] huh, entered softirq 2 NET_TX ffffffff81522420 preempt_count 00000200, exited with 00000201?
[131856.147773] huh, entered softirq 3 NET_RX ffffffff81524460 preempt_count 00000200, exited with 00000bfe?
[131856.147781] huh, entered softirq 8 HRTIMER ffffffff8105ff60 preempt_count 00000200, exited with 000001ff?
[131856.147787] huh, entered softirq 9 RCU ffffffff810a6a70 preempt_count 00000200, exited with 000001fc?
[131856.147792] huh, entered softirq 3 NET_RX ffffffff81524460 preempt_count 00000201, exited with 00000201?
Oracle Corp.
Re: Virtualbox-4.3.0 - multiple hard crashes
I'll give it a try and report back, most likely it will.skoehler wrote:So it's a linux guest? There is a tool called "stress". It allows you to generate plenty of CPU and I/O load. Can you reproduce the crash with that?saintdev wrote:When I update the Arch guest.
All I have to do is run pacman -Syu and the guest is not even able to finish updating before causing a kernel panic on the host.
Re: Virtualbox-4.3.0 - multiple hard crashes
Do you want the guest (Arch) or host (Gentoo) config? I already posted the host's config in my previous post (except I enabled frame pointers for the second kernel log).Ramshankar wrote:Could you post "cat /boot/config-`uname -r`" or whatever the equivalent is on Arch linux.
Edit: Also, would testing a kernel with preemption disabled help you in any way?
-
- Oracle Corporation
- Posts: 793
- Joined: 7. Jan 2008, 16:17
Re: Virtualbox-4.3.0 - multiple hard crashes
No, the host kernel config. is enough. Thanks. I can't yet say what exactly goes wrong with it, whether CONFIG_PREEMPT_RCU or CONFIG_PREEMPT itself is causing problems. Here's what a stock Ubuntu kernel has which doesn't, afaict, have such lockups:saintdev wrote:Do you want the guest (Arch) or host (Gentoo) config? I already posted the host's config in my previous post (except I enabled frame pointers for the second kernel log).Ramshankar wrote:Could you post "cat /boot/config-`uname -r`" or whatever the equivalent is on Arch linux.
Edit: Also, would testing a kernel with preemption disabled help you in any way?
Code: Select all
# CONFIG_PREEMPT_RCU is not set
CONFIG_PREEMPT_NOTIFIERS=y
# CONFIG_PREEMPT_NONE is not set
CONFIG_PREEMPT_VOLUNTARY=y
# CONFIG_PREEMPT is not set
Oracle Corp.
-
- Posts: 54
- Joined: 14. Feb 2010, 22:34
- Primary OS: Linux other
- VBox Version: OSE self-compiled
- Guest OSses: Win 7, Win XP, Gentoo 32-bit
Re: Virtualbox-4.3.0 - multiple hard crashes
For the sake of conversation, this is what I have in my 3.10.18 gentoo-sources kernel:Ramshankar wrote:No, the host kernel config. is enough. Thanks. I can't yet say what exactly goes wrong with it, whether CONFIG_PREEMPT_RCU or CONFIG_PREEMPT itself is causing problems. Here's what a stock Ubuntu kernel has which doesn't, afaict, have such lockups:saintdev wrote:Do you want the guest (Arch) or host (Gentoo) config? I already posted the host's config in my previous post (except I enabled frame pointers for the second kernel log).Ramshankar wrote:Could you post "cat /boot/config-`uname -r`" or whatever the equivalent is on Arch linux.
Edit: Also, would testing a kernel with preemption disabled help you in any way?Again, it's just something to narrow down the problem a little bit than what is known now.Code: Select all
# CONFIG_PREEMPT_RCU is not set CONFIG_PREEMPT_NOTIFIERS=y # CONFIG_PREEMPT_NONE is not set CONFIG_PREEMPT_VOLUNTARY=y # CONFIG_PREEMPT is not set
mark@c2RAID6 ~ $ zcat /proc/config.gz | grep PREEMPT
CONFIG_TREE_PREEMPT_RCU=y
CONFIG_PREEMPT_RCU=y
CONFIG_PREEMPT_NOTIFIERS=y
# CONFIG_PREEMPT_NONE is not set
# CONFIG_PREEMPT_VOLUNTARY is not set
CONFIG_PREEMPT=y
CONFIG_PREEMPT_COUNT=y
CONFIG_DEBUG_PREEMPT=y
mark@c2RAID6 ~ $
As with everything Gentoo we're sort of on our own. I have no reason to trust these settings, assuming they are involved at all. However this crashes with 4.3.x and works fine with 4.2.18. I am willing to try some changes if there's a reasonable willingness to evaluate going forward.
I do currently have more kernel debugging turned on than I used to.
- Mark
Re: Virtualbox-4.3.0 - multiple hard crashes
the CONFIG_PREEMPT_* settings are controlled by the option "Preemption Model" under "Processor Type and Features" in the kernel's menuconfig. I switched from "Preemptible Kernel (Low-latency Desktop)" to "Voluntary Kernel Preemption (Desktop)". The CONFIG_PREEMPT_* settings now match what Ramshankar posted above about the Ubuntu kernel.
I will let you know, whether I experience the host crash with virtualbox 4.3 again.
I will let you know, whether I experience the host crash with virtualbox 4.3 again.
Re: Virtualbox-4.3.0 - multiple hard crashes
I also made these changes and I can report that I'm not having any problems.skoehler wrote:the CONFIG_PREEMPT_* settings are controlled by the option "Preemption Model" under "Processor Type and Features" in the kernel's menuconfig. I switched from "Preemptible Kernel (Low-latency Desktop)" to "Voluntary Kernel Preemption (Desktop)". The CONFIG_PREEMPT_* settings now match what Ramshankar posted above about the Ubuntu kernel.
I will let you know, whether I experience the host crash with virtualbox 4.3 again.
My host is 32 bit Gentoo/linux. Guest is Windows 7 Professional 64-bits.
Before this change the host would crash during the boot or login procedure. It has been working for over 30 minutes without any problem.
Re: Virtualbox-4.3.0 - multiple hard crashes
Update for my case:Perryg wrote:I would upgrade to 4.3.2 myself.web-devel wrote:Last week I upgraded to Suse 12.3/Virtualbox 4.3.0 and feel I now have a very similar problem. Hard freeze on the base Suse OS. The console is sadly unresponsive. Turning VM to use only one processor seems to help. I'd love to downgrade but a package < 4.3.0 does not appear to be available for Suse 12.3.
I Downgraded to version 4.2.14. from this location.
http://forums.opensuse.org/english/get- ... 2-a-2.html and have no problems since.
-
- Posts: 54
- Joined: 14. Feb 2010, 22:34
- Primary OS: Linux other
- VBox Version: OSE self-compiled
- Guest OSses: Win 7, Win XP, Gentoo 32-bit
Re: Virtualbox-4.3.0 - multiple hard crashes
As of yesterday I recompiled with the more modest preempt settings:skoehler wrote:the CONFIG_PREEMPT_* settings are controlled by the option "Preemption Model" under "Processor Type and Features" in the kernel's menuconfig. I switched from "Preemptible Kernel (Low-latency Desktop)" to "Voluntary Kernel Preemption (Desktop)". The CONFIG_PREEMPT_* settings now match what Ramshankar posted above about the Ubuntu kernel.
I will let you know, whether I experience the host crash with virtualbox 4.3 again.
mark@c2RAID6 ~ $ zcat /proc/config.gz | grep PREEMPT
# CONFIG_PREEMPT_RCU is not set
CONFIG_PREEMPT_NOTIFIERS=y
# CONFIG_PREEMPT_NONE is not set
CONFIG_PREEMPT_VOLUNTARY=y
# CONFIG_PREEMPT is not set
mark@c2RAID6 ~ $
and then updated to virtualbox-4.3.2 under Gentoo. I did not update the guest additions immediately but rather ran 2 Win 7 VMs along with the VMWare Player VM with no problems. In that state I've run about 12 hours without problems. I don't think I ever went more than 1-2 hours with the other preempt settings & 4.3.x so these setting changes appear to be a reasonable workaround. (But only a work around - NOT a solution...)
I've just this morning updated the guest additions on both VMs and am now working with that setup. I'll continue to track this over the next few days and report back later.
I will say that these settings do seem to have slowed things down. The machine doesn't feel quite as fast but that may be imaginary as I haven't run anything other than the low latency desktop setting for years.
I do hope the Virtualbox developers will continue to work on this issue.