PCI-passthrough fails with Ubuntui kernel 3.5 and above

[ucrasher]

Hi, was using PCI passthrough since 1.5 years. Host Ubuntu 12.04 with kernel 3.2.x, VB different versions (4.x to 4.3.4). After upgrade to raring kernel the PCI-passthrough fails. Trying kernel 3.5.x showed the same. The PCI device is one of the USB controllers on the mainboard. The fail situation prevents the windows7 in guest to start the root hub driver (error 43). BTW: USB passthrough is not an option to what I need to achieve. It is too slow if USB devices re-enumerate after a firmware flash, so I cannot debug firmware flashers.

Attachments: syslog and VB log for both kernels, and guest.vbox. I can do tests if you need that.

[ucrasher]

Hi, tested with saucy kernel (3.11), still does not work. I had hoped for some improvements with that kernel since the change log contains quite some lines talking about IOMMU fixes.

If there is anything I can do myself, please let me know.

PS: If you are still in Endersbach, then I can bring the machine for some test if you need that.

[ucrasher]

​https://forums.virtualbox.org/viewtopic.php?f=7&t=57200

Now using Ubuntu 14.04 with kernel 4.2 from Wily. VB is 5.016.

Syslog

vboxpci 0000:00:1d.0: reg=0 start=f7f17000 size=1024
DMAR: DRHD: handling fault status reg 3
DMAR: DMAR:[DMA Read] Request device [00:1d.0] fault addr 697e2000 
DMAR:[fault reason 06] PTE Read access is not set

With a Linux guest, I see the device with lspci, but it probably does not get interrupts. My guess is that the above failure prevents them to get passed into the VM.

[ucrasher2]

portion of syslog showing failed IOMMU calls

[ucrasher2]

vbox.log for VB 5.1.6 and kernel 4.4 on Ubuntu 16.04

[ucrasher2]

Hi, I attached a few new files from VB 5.1.6, Ubuntu 16.04 with kernel 4.4. There are quite still some error lines in syslog.

Example

DMAR: DRHD: handling fault status reg 2
DMAR: DMAR:[DMA Read] Request device [02:00.0] fault addr 8d910000
DMAR:[fault reason 06] PTE Read access is not set

And at termination of the VM, it even crashes with "kernel NULL pointer dereference". This can be seen in the attached syslog.