esxi 5.5 update1 crash (purple screen of death) on PCIe passthrough

[budric]

Hi,

If I create an Ubuntu VM with PCIe passthrough for an onboard graphics card, the whole ESXi host crashes when I start up the machine.  This happens for Ubuntu 13.10 and 12.04.

The same configuration works fine on Windows 7.  I can install the driver and make full use of the card in PCIe passthrough mode.  The card runs at the speed that it should for days on end.  I do not believe it's faulty hardware.

Anyone experience anything similar?  How do I fix it?

Where are the crash logs kept on esxi?  I can ssh to it but i don't know where to look.

Thanks.

[grasshopper]

Hi budric,

Following a PSoD you should try to take a screenshot of the diagnostics data on the screen (or at least camera phone pic).  Next, powercyle the system then gather a vm-support log (could be useful if you need to engage official VMware support.  It also contains all relevant data if interested in unzipping and digging in).

On a properly configured system the crash data should have been dumped off for later review (probably in /var/core/).  You can use the 'esxcfg-dumppart -L vmkernel-zdump-filename' command to extract the human readable info (will be near the end of the file, funny enough after the word @BlueScreen).  Unfortunately, YMMV with passthrough of on-board devices.  For best results use a PCIe add-in card.

[budric]

Thanks,

I'm running free version of ESXi 5.5 to evaluate it for our needs, so I don't know how far I'll get with VMware support.  Purple screen says it's a hardware problem - well if that was the case it wouldn't run in Windows VM at all.  But if I configure a windows VM with same card, everything works.  Problem manifests itself on ubuntu only, and it crashes the whole VMWare host.

I was wondering if it's just me having this issue or if it's common?  Anybody able to get Linux VM to run with PCIe passthrough graphics card?

[seghu]

We have the same issue, it sometimes happens at boot up.

but happens everytime on reboot.

@

[BujarN]

I have the same problem here on HP Proliant ML350 Gen8. Configured a hp smart array P421 for passthrough and then added the smart array to a sever 2012 r2 VM, and the VM did not boot and then it caused the ESXi host to crash and get the same purple screen.