Solaris 64-bit guests pegging core @ 100% after attempting to log in

[Peter Jorgensen]

I have a couple of 64-bit Solaris 10 VMs (one Update 3, one Update 4) installed on an OpenSolaris 2008.11 host. If I just boot the VMs up and leave them for a couple of hours (sometimes less; the time seems to vary), when I attempt to log in to them (ssh or 'console') they 'lock up' and each one pegs one of my Sun Ultra 20's cores @ 100%. The VMs were doing nothing but sitting idle at the time.

I have observed this with VirtualBox 2.1.2, 2.1.0, 2.0.6... I have tried starting with VMs with and without VRDP enabled. I have tried starting with VMs with and without using VBoxHeadless (both with and without VRDP enabled). I have observed the same behaviour with the host running Solaris 10 update 5 (5/08). I have observed this with and without the VBox guest additions installed. I have observed this with only 1 VM running or both. Each is configured for 1.6 GB of RAM on my 4GB host.

I have attached the log file as well as pstack and a few seconds of truss output from when the VBox process is pegging the core.

Incidentally, a colleague has been running a 32-bit Windows XP guest on his Solaris 10 Update 5 Ultra 20 for a couple of months straight without issue (VBox 2.0.6, I believe). His Operton CPU is older does not have AMD-v.

[Peter Jorgensen]

pstack and truss output

[Frank Mehnert]

Please reopen if this problem persists with the latest release 2.2.2. There were some improvements.

[Peter Jorgensen]

Problem appears to be fixed in 2.2.2. Thanks very much!

[Peter Jorgensen]

I spoke too soon! I just tried running a 64-bit Solaris 10 update 6 (10/08) guest on my Solaris 10 update 6 host and after about 3 hours of me just letting it sit there and do nothing, I tried hitting return in the console window and it is now running away with one of my cores.

Attachments to follow...

[Peter Jorgensen]

The Machine->Show Log output

[Peter Jorgensen]

pstack and truss output

[Peter Jorgensen]

Oh, and I am using VBox 2.2.2 (but I guess you could see that from the Show Log output).

[rasta]

This problem appears to be the same as ticket #2258. If it is, someone should consolidate them. The main issue is unstable behavior and often pegged core when using 64-bit Solaris 10 guests.