raw vmdk disks cause ata timeouts on guest

[Keith]

After upgrading to 4.0.6 today, a freebsd VM is no longer able to access four vmdk disks that were created by createrawvmdk command. I tried recreating the vmdk's and attached the new ones to the guest but still receive ATA timeouts on boot. The guest OS detects the sata controller and disks correctly.

I've attached a Vbox.log and an older vbox.log from 4.0.4, when the raw disks functioned perfectly.

[Keith]

Log file from 4.0.6

[Keith]

Vbox.log from 4.0.4

[Keith]

Not sure if this info will help -

The guest can access up to two raw vmdk disks without any indication of a problem. Throughput (tested with dd) seems fine. Once I storageattach a third raw vmdk, I get ATA timeout errors on the guest and the disks are not accessible.

[Keith]

This is no longer a problem after an upgrade to 4.0.8 and recompiling the kernel modules with dkms. I think I may have caused the problem by having my kernel modules out of sync with the version of virtualbox.

[Keith]

I rebooted my vm server after 72 day uptime and the problem returned while running 4.0.8. I then upgraded to 4.0.10, recompiled modules w/ dkms, rebooted and the problem still exists. FreeBSD is reporting the following errors: ad6: TIMEOUT - FLUSHCACHE retrying (1 retry left) ad6: FAILURE - FLUSHCACHE timed out

On three of my four raw vmdk disks. It looks like the problem is solved by turning on the hosts io caching: vboxmanage storagectl freenas2 --name "SATA Controller1" --hostiocache on

[aeichner]

No activity in 9 years with a lot of storage related changes inbetween. Closing as obsolete.