USB Passthrough and my Bad Mojo

[iphillips77]

I'm so tantalizingly close to being able to put my unraid server back down in the basement where it belongs..  If only I could get a working USB solution for my two (1x Win7, 1x Mac OS X) VMs!

 

I posted last week here http://lime-technology.com/forum/index.php?topic=43816.msg441303#msg441303 asking for any advice on my Mac-side USB problems, that are almost certainly a driver issue inside the VM.  In the meantime, I bought the bullet and purchased an advertised Mac-compatible PCIe card, and it looked like it had solved all my problems..  Oh boy was I wrong!

 

The card, an Orinoco 4-port Fresco Logic based card, passes through and works properly on the Mac side.  But when I restart the Mac VM from inside the VM, the whole server locks up.  Like, hard, refuses telnet connections locks up.

 

If I just shut down the Mac VM, either from inside or doing a force stop from unraid's webui, the server continues to run normally, but I can't start that VM again.  It's unable to pass through the USB card a second time.  And what's weirder, rebooting unraid does NOT allow me to start the VM again either.  I have to actually hard reboot, power down the system and bring it up again cold.

 

Here's some things I saw in my syslog pertaining to the card.  I'm guessing it's just bad firmware, card-won't-work kinda stuff that means I'm out $30 and have to just buy a different one.  But hoping there's something I can try to coax it into working.

 

Feb  8 06:34:48 Behemoth kernel: pci 0000:09:00.0: [1b73:1100] type 00 class 0x0c0330
Feb  8 06:34:48 Behemoth kernel: pci 0000:09:00.0: reg 0x10: [mem 0xfb100000-0xfb10ffff 64bit]
Feb  8 06:34:48 Behemoth kernel: pci 0000:09:00.0: reg 0x18: [mem 0xfb111000-0xfb111fff 64bit]
Feb  8 06:34:48 Behemoth kernel: pci 0000:09:00.0: reg 0x20: [mem 0xfb110000-0xfb110fff 64bit]
Feb  8 06:34:48 Behemoth kernel: pci 0000:09:00.0: supports D1
Feb  8 06:34:48 Behemoth kernel: pci 0000:09:00.0: PME# supported from D0 D1 D3hot

 

So far, so good?

 

Feb  8 06:34:48 Behemoth kernel: dmar: [Firmware Bug]: RMRR entry for device 09:00.0 is broken - applying workaround

 

.....oh, maybe not.  This line makes me think that maybe the card is buggered.

 

Feb  8 06:34:48 Behemoth kernel: IOMMU: Setting identity map for device 0000:09:00.0 [0x36e33000 - 0x36e41fff]
Feb  8 06:34:48 Behemoth kernel: pci 0000:09:00.0: Signaling PME through PCIe PME interrupt
Feb  8 06:34:48 Behemoth kernel: pci-stub 0000:09:00.0: claimed by stub

 

(Yeah, I tried adding it as a pci-stub.id entry in syslinux.cfg to see if that would help.  It didn't.)

 

Now, when shutting down the VM, I spot this little nugget.

 

Feb  8 06:42:50 Behemoth kernel: vfio-pci 0000:09:00.0: timed out waiting for pending transaction; performing function level reset anyway

 

Uh oh!  That doesn't look good.  Trying to restart the VM from this point gives us these lines pertaining to the card.

 

Feb  8 15:44:01 Behemoth kernel: vfio-pci 0000:09:00.0: Refused to change power state, currently in D3
Feb  8 15:44:01 Behemoth kernel: vfio-pci 0000:09:00.0: Refused to change power state, currently in D3
Feb  8 15:44:01 Behemoth kernel: vfio-pci 0000:09:00.0: Refused to change power state, currently in D3
Feb  8 15:44:02 Behemoth kernel: vfio-pci 0000:09:00.0: timed out waiting for pending transaction; performing function level reset anyway
Feb  8 15:44:02 Behemoth kernel: vfio_cap_init: 0000:09:00.0 hiding cap 0xff
Feb  8 15:44:02 Behemoth kernel: vfio_cap_init: 0000:09:00.0 hiding cap 0xff
Feb  8 15:44:02 Behemoth kernel: vfio_cap_init: 0000:09:00.0 hiding cap 0xff
......(repeats 50 more times).......
Feb  8 15:44:02 Behemoth kernel: vfio_cap_init: 0000:09:00.0 hiding cap 0xff
Feb  8 15:44:02 Behemoth kernel: vfio_cap_init: 0000:09:00.0 hiding cap 0xff
Feb  8 15:44:02 Behemoth kernel: vfio_ecap_init: 0000:09:00.0 hiding ecap 0xffff@0x100
Feb  8 15:44:02 Behemoth kernel: vfio_ecap_init: 0000:09:00.0 hiding ecap 0xffff@0xffc
......(same with the ecap bits)......
Feb  8 15:44:02 Behemoth kernel: vfio_ecap_init: 0000:09:00.0 hiding ecap 0xffff@0xffc
Feb  8 15:44:02 Behemoth kernel: vfio_ecap_init: 0000:09:00.0 hiding ecap 0xffff@0xffc
Feb  8 15:44:02 Behemoth kernel: vfio_ecap_init: 0000:09:00.0 hiding ecap 0xffff@0xffc
Feb  8 15:44:02 Behemoth kernel: genirq: Flags mismatch irq 18. 00000000 (vfio-intx(0000:09:00.0)) vs. 00000080 (ehci_hcd:usb1)
Feb  8 15:44:03 Behemoth kernel: vfio-pci 0000:09:00.0: Refused to change power state, currently in D3
Feb  8 15:44:04 Behemoth kernel: vfio-pci 0000:09:00.0: timed out waiting for pending transaction; performing function level reset anyway

 

And then this pops up in unraid.

 

internal error: early end of file from monitor: possible problem:
2016-02-08T20:44:02.800332Z qemu-system-x86_64: -device vfio-pci,host=09:00.0,bus=root.1,addr=00.2: vfio: Error: Failed to setup INTx fd: Device or resource busy
2016-02-08T20:44:02.801472Z qemu-system-x86_64: -device vfio-pci,host=09:00.0,bus=root.1,addr=00.2: Device initialization failed
2016-02-08T20:44:02.801500Z qemu-system-x86_64: -device vfio-pci,host=09:00.0,bus=root.1,addr=00.2: Device 'vfio-pci' could not be initialized

 

Any insights?

[saarg]

Your USB card most likely doesn't support resetting. That is why you see the INTx message.

There is not much you can do about it other than buying a card that support INTx.