After Updating to 5.0.2r102096 Windows Guest hangs

[bitkocher]

After Update to 5.0.2r102096 (Fedora 22 64Bit host) Windows 10 hangs on boot. Last lines from log are:

00:00:07.250639 AioMgr1-N: Request 0x007f4944001d80 failed with rc=VERR_INVALID_HANDLE, migrating endpoint /dev/vg_big/sr71win7 to failsafe manager.
00:00:07.250742 AIOMgr: I/O manager 0x007f493a13efe0 encountered a critical error (rc=VERR_FILE_AIO_NO_REQUEST) during operation. Falling back to failsafe mode. Expect reduced performance
00:00:07.250755 AIOMgr: Error happened in /home/vbox/vbox-5.0.2/src/VBox/VMM/VMMR3/PDMAsyncCompletionFileNormal.cpp:(1664){int pdmacFileAioMgrNormal(RTTHREAD, void*)}
00:00:07.250761 AIOMgr: Please contact the product vendor

Same for Windows 7.

It seems 102096 is really broken on Fedora 22.

After downgrade to 101573 everything works again.

[noteirak]

Such reproducable error should be posted on the BugTracker. Check if a ticket already exists or create a new one if not.
In both cases, make sure to put a FULL VM log file of the session reflecting the issue. You can post the ticket ID here afterwards for follow up.

[Perryg]

The DEVs really don't like compressed log files unless they are too big to post normally. More work for them.

[noteirak]

Amended

[BezaiServices]

This is install is on a Windows 10 home upgrade.
Vbox is ver 5.0.2
The W8 Enterprise hangs.
After booting, the 64bit guest options disappear.

[eworm]

Looks like Virtualbox 5.0.2 fails if host cache is not used. Edit virtual machine, got to "storage" and enable "User Host I/O Cache".

[Dani3lSun]

Just for info.
I had the same issue today on my debian wheezy prod machine. Updated today from 4.3.28 to 5.0.2 and every linux VM failed to start with a disk error (VERR_INVALID_HANDLE).
Enabling Host Cache for the SATA controller solved the issues. Hope the VMs run stable now, because on 4.3 I had to disable Host Cache because the VMs aborted at night then backups ran with more IO load.
My VBox disks run on LVM Raw Images, when taking LVM snapshot backups the VMs aborted...Hopefully this behaviour is not present with version 5...