Ticket #8744 (closed defect: fixed)
linux kernel >2.6.32 grinds to a halt after a while
| Reported by: | yrro | Owned by: | |
|---|---|---|---|
| Component: | other | Version: | VirtualBox 4.0.4 |
| Keywords: | Cc: | ||
| Guest type: | Linux | Host type: | Windows |
Description
I have found that booting into any kernel later than 2.6.32 will result in the VM hanging after a period of time (this can be as short as 15 minutes or a long as a couple of hours).
When this occurs the symptoms are that the keyboard may no longer be used to type into gnome-terminal, though I can still type into other programs. Attempting to log out or reboot the machine via the GNOME gui has no effect. As this occurs, more and more programs stop responding, until the only way out is a hard reboot.
This is 100% reproducible after booting into 2.6.38. I think I saw it with 2.6.37 before I upgraded to 2.6.38 though I am less sure. 2.6.32 works fine.
Attachments
-
wintermute-2011-04-17-22-20-13.log
(58.2 KB) -
added by yrro 9 years ago.
- virtualbox log
-
ttyS0t
(75.5 KB) -
added by yrro 9 years ago.
- linux kernel log messages
Change History
Changed 9 years ago by yrro
-
attachment
wintermute-2011-04-17-22-20-13.log
added
comment:1 Changed 9 years ago by yrro
In the above bug I reset the VM at 1:05:57. Sadly there doesn't appear to be anything useful in the log, nor is there anything useful logged by the kernel.
comment:2 Changed 9 years ago by frank
Could you also attach the output of "dmesg" from the guest? I would prefer if you could boot your guest kernel with the following kernel parameters attached:
ignore_loglevel apic=debug console=ttyS0
Then make sure to enable the 1st COM port for that VM and select "file" as mode. Then wait until your guest halts. Then attach the resulting file to this ticket. I hope the guest kernel writes something useful to the log when this happens ...
comment:3 Changed 9 years ago by yrro
Unfortunately the kernel doesn't appear to log anything useful. :(
comment:4 Changed 9 years ago by yrro
In case this is relevant, I've observed this problem with 4.0.8 with two different CPUs: a Core i7 920 (running Debian) and a core i5 950 (running Fedora 15).
comment:5 Changed 9 years ago by yrro
I've found that reducing the number of CPUs in my VMs to 1 has worked around this problem. I have left the IO APIC support enabled.
comment:7 Changed 9 years ago by yrro
I have not seen this since upgrading to VirtualBox 4.1.0, on Linux 2.6.39.
I now run VirtualBox 4.1.2 and Linux 3.0.0 and everything is working fine.
virtualbox log