Troubleshooting ESXi hardware freeze with PCI Passthrough

Question

I have a Supermicro X9SCM board with a Atheros AR5008 PCI Express card (D-LINK DWA-556, Device=0024&Vendor=168C). The card can successfully be marked for PCI passthrough in ESXi (I tried versions 4.1 and 5.0) though each time I start a VM with the Wifi card associated - the entire host freezes and requires a hard reset.

There is a good chance this card is just not compatible for some reason - though there seems to be at least one report I could find of it "working", or at least the guest being able to boot. I would really like to understand why it is failing though. I have tried digging into some log files and other resources to see if I can glean any knowledge on how to best troubleshoot this, though I am far from an expert with VMWare tools.

Here is what I have looked at so far:

• BIOS, tried latest version (1.1a) and one older version (1.0c).
• The BIOS has a log that reports "PCI ERR" or "PCI ERR - Asserted" whenever this freeze event happens.
• I grabbed the various logs from /var/log on the ESXi host, though I haven't really been able to see anything too useful from them just yet. Maybe I don't know where to look.
• I tried adding the PCI card to the passthru.map file to see if I could perhaps hint to ESXi how it should behave with no luck. (Note: I haven't tried all combinations of reset method / fpt shareable yet)
• I have read there may be a difference with "Active" PCI Express cards. I believe this might be referring to Active State Power Management though I am not sure how to even check this.
• I have contacted Supermicro support to see if there is a known issue with the BIOS / hardware though I haven't heard back. I also tried to get on the VMWare communities and post on their forums though I haven't been able to activate my account for some strange reason.

Again, my real question is: How do I go about understanding why this device is causing the hypervisor to lockup when it is assigned to a guest?

Solution 1

It likely is not the Hypervisor that is locking up, but some kind of hardware (like the PCIe switch). You would have a hard time debugging this without any kind of PCIe debugging hardware and a whole bunch of PCIe-specific knowledge, so it probably is not worth pursuing. In general, PCI passthrough is not what you should use without a great deal of consideration.

If you need a wireless-connected interface on the virtual machine, consider using an external device (router/bridge) bridging the wireless network to a wired one and using a virtual interface from within your virtual machine connecting to this network. Another option would be using a USB-plugged interface together with a USB network redirector.

Solution 2

Realize this question might be dated but still came up as a top google result when I hit this problem. Found a solution here so figured I'd share.

My env:

• VMware ESXi 6.5
• SuperMicro SYS-E300-8D
• OPNsense (FreeBSD based) guest
• Qualcomm Atheros AR9462

Assumptions:

• You've physically installed the card.
• You've enabled it for passthrough in ESXi and rebooted
• You've added this PCI device to a VM

Answer:

Set the correct passthrough params in the /etc/vmware/passthru.map adding an entry like so.

# passthrough attributes for devices
# file format: vendor-id device-id resetMethod fptShareable
# vendor/device id: xxxx (in hex) (ffff can be used for wildchar match)
# reset methods: flr, d3d0, link, bridge, default
# fptShareable: true/default, false

# Atheros Wireless
168c  0034  d3d0     false

Then reboot the host.

Note 1: The first and second fields here are vendor and device specific but I got the values (already in hex) from the ESXi web interface under Host>Manage>Hardware>PCI Devices

Note 2: There's other useful stuff in the passthru.map already so don't just overwrite, append.

I also set

pciPassthru0.msiEnabled=false

in the .vmx file for the guest but before getting the passthru.map fixed so I'm not positive it's necessary.

I haven't finished configuring the card or making it useful but my host now stays running when the VM starts and the guest detects it.

Solution 3

I'm curious as to what your plan is here. You're trying to run ESXi PCI passthrough of a wireless NIC to a guest virtual machine? What is the guest OS? What is the purpose of this setup?

Not all PCIe devices are compatible with VMDirectPath. This may simply be one of them.

Comments

[JSchlather]

I am trying to pass it through to a pfSense VM

[Chopper3]

@ewwhite - couldn't agree more, there's a limited number of VMDP capable adapters. Not ever card supports VT-d or IOMM, even then you need a supporting BIOS/EUFI. That's why people usually pick supported cards in the first place.