Guest almost freezes (100% CPU load) when trying to access disk - nothing written to file.

[Ingo]

This freezes I am observing with every single version of VBox > 4.0.4.

I usually have a VM with OS/2 4.5 up and running fine daily. This VM almost never gets shut down, it is just saved and also survives 100's of Suspends of the host (s2ram). Starting with VBox 4.0.6 the VM worked fine onlyy for some 10 save/s2ram cycles, then suddenly CPU load in the VM gous up to 100% and the "green dot" at the bottom of the VBox-window indicates continous disk (read)-access to the virtual disk, however host indicates zero disk/file activity.

I finally set up a separate test system with Squeeze-amd64 and VBox 4.1.0 and a known good VDI attached as immutable. Yesterday the above described situation happened: first mouse clicks get delayed and finally lost, CPU-load up at 100%, disk activity 100% in VBox window (bottom). I saved this situation and do attach several log files:

a) VBox-frozen.log (Guest-log where VM started working, then froze at the end, saved VM)
b) VBox-starting-frozen-VM.log (opening the saved + frozen VM today, still the same, saved again)
c) VirtualBox-home.tgz (all global configuration files including logs)

Important thes global log files contain lot of ERROR entries like:

00:00:04.438 nspr-2   ERROR [COM]: aRC=VBOX_E_INVALID_VM_STATE (0x80bb0002) aIID={5eaa9319-62fc-4b0a-843c-0cb1940f8a91} aComponent={Machine} aText={Machine is not locked for session (session state: Unlocked)}, preserve=false
00:08:27.310 nspr-2   ERROR [COM]: aRC=VBOX_E_IPRT_ERROR (0x80bb0005) aIID={5eaa9319-62fc-4b0a-843c-0cb1940f8a91} aComponent={Machine} aText={Saved screenshot thumbnail data is not available (VERR_NOT_SUPPORTED)}, preserve=false
00:08:37.453 main     ERROR [COM]: aRC=VBOX_E_OBJECT_IN_USE (0x80bb000c) aIID={53f9cc0c-e0fd-40a5-a404
....

As mentioned above the VM has an immutable VDI file and one snapshot.

[Ingo]

guest log with VM almost frozen at the end. VM saved.

[Ingo]

guest log starting a saved and frozen VM, closing by "save" again

[Ingo]

global VBox configuration files incl. logs wit lot of ERROS's

[Ingo]

Remark:

I have a dedicated installation with said configuration. If further tests or information is needed, please advise.

[Ingo]

global (~/.VirtualBox) VBox configuration files incl. logs wit lot of ERROS's

[Ingo]

Added archive of global (~/.VirtualBox) configuration files incl. logs with lot of ERROS's a second time - now just *.zip. For some unknown reason the *.tgz file got broken during up/down-load.

[Ingo]

For completeness here all differential disk files for the "frozen machine": In my opinion at least 1 more than needed for 1 snapshot (taken in saved state) and closed/saved?

root@pp:/home/ingo# ls -l /home/ingo/VirtualBox\ VMs/OS2/Snapshots/
insgesamt 69384
-rw------- 1 ingo ingo 28659952 20. Jul 22:06 2011-07-20T20-06-13-310805000Z.sav
-rw------- 1 ingo ingo 32797486 13. Aug 18:14 2011-08-13T16-14-50-208722000Z.sav
-rw------- 1 ingo ingo  1069056 20. Jul 22:06 {381bc69c-ee8d-4bd9-93d6-68f07c27152f}.vdi
-rw------- 1 ingo ingo    20480 10. Jul 19:51 {88894f99-344f-480b-b8c3-b287b5ad0010}.vdi
-rw------- 1 ingo ingo  8409088 13. Aug 18:14 {d4b10aa1-6632-407e-b360-3bc754eab852}.vdi

Machine was created under VBox 4.1.0 with main *.vdi attached as immutable.

[Ingo]

After extensive tests I mark this fixed in VBox 4.1.4.

However it would be nice to know more details of the root cause or corresponding fix, as 4.1.4 is the first one in which stability is back to 3.2.12 level.

[Frank Mehnert]

ingo2, this is hard to say. Are you sure that this bug was fixed in VBox 4.1.4 or did you just have other issues running VBox 4.1.0 and 4.1.2?

[Ingo]

frank,

thanks for your attention. I am really sure that the bug was still present in 4.1.0 (the version I did report it for). I am also sure that it showed up in 4.0.6 and every subsequent version through 4.0.12. I am not 100% sure that it was still present in 4.1.2. I had installed 4.1.2 on a test system on a extra partition and only observed intermittent mouse problems, but not *.vdi corruption. As 4.1.4 beheaved fault free on the test system, I decided to upgrade 4.0.4 -> 4.1.4 on my work system and everything is absolutely reliable.

Remark: the critical guest system for this bug was OS/2 Warp 4.52

[aeichner]

Please reopen if still relevant with a recent VirtualBox release.