[Issue] bfm: ioctl failed: `IOCTL_START_VMM`

[no-realm]

Hi,

two of my friends seem to have issues with starting the hypervisor.
It loads (make load) just fine, but when they try to start (make start) it, the get an error:

bfm: ioctl failed: `IOCTL_START_VMM`

First I though that it was just a single case, but now another friend of mine encountered the same issue.
For me it works just fine, so I don't get why it won't work for them.
Both of them use Windows 10.

[rianquinn]

Check via serial to see what message came up. Likely it's something simple like virtualization extensions are not turned on.

[no-realm]

Serial?

[rianquinn]

Bareflank supports serial output in addition to make dump

[no-realm]

Eh, how do I do that?
I tried PortMon, but that didn't let me capture anything.

[rianquinn]

You need to identify which port your connected to and then tell Bareflank to use that port:
http://bareflank.github.io/hypervisor/html/index.html#serial

[nick-kvmhv]

My situation must be related to Windows 10 itself: I have no problem starting it on Ubuntu on the same laptop. Windows 10 gives me identical error when I boot in in KVM or as the main OS. (for KVM I enabled nested virtualization, although I am not sure how well it works compared to vmware). If there is more output in serial than in DebugView, I'll try to dust off my old serial over USB cable and check.

[rianquinn]

Yeah, you should get more input as it died safely which means that it knows what the issue is, you just need to get access to the error. Make dump only works when the vCPUs are up and running.

[nick-kvmhv]

Before I figure out how to connect that serial cable to a laptop with no serial ports, here is what I get in DbgView when I try to start it:
[BAREFLANK ERROR]: start_vmm failed
[BAREFLANK ERROR]: IOCTL_START_VMM: ioctl_start_vmm failed: 8000000000000030 - ENTRY_ERROR_VMM_START_FAILED
I can turn Hyper V on with no issues.

[rianquinn]

yeah that just means that it attempted to start but failed. Once option would be to create a global debug ring and only dump from the debug ring, in which case you could get the output from make dump without needing serial

[nick-kvmhv]

When running it under KVM, I got this in COM1 output when starting:

---

• Standard Exception Caught -

---

vector

[rianquinn]

Huh... that's weird. that usually means that std::vector generated an exception, likely due to out of bounds. @connojd Have you seen this with KVM? Also... what hardware are you running on?

[nick-kvmhv]

I have this: Intel(R) Core(TM) i7-6600U CPU @ 2.60GHz on Lenovo x260 laptop. Keep in mind that It works like a charm under Ubuntu on the same laptop. I have W10 on a SSD USB that I use to boot both directly and under KVM and it gives me this error. I'll be happy to re-compile and re-run with any added instrumentation you can suggest.

[rianquinn]

I will take a look at this on Monday. I have a skylake machine I can test Windows on just to make sure something is broken. @ktemkin do you mind if @connojd takes a quick look at KVM next week?

[ktemkin]

That's fine-- @connojd, feel free to spend a bit of time looking into this. :)

[nick-kvmhv]

A quick update: it seems to be failing in
vmcs_intel_x64::write_32bit_guest_state
on the following line:
vmcs::guest_fs_limit::set(state->fs_limit());

I added state->dump(); call before it and here is what I got:

DEBUG: ----------------------------------------
DEBUG: - vmcs_intel_x64_host_vm_state dump    -
DEBUG: ----------------------------------------
DEBUG: 
DEBUG: segment selectors:
DEBUG:     - m_es: 0x2b
DEBUG:     - m_cs: 0x10
DEBUG:     - m_ss: 0x18
DEBUG:     - m_ds: 0x2b
DEBUG:     - m_fs: 0x53
DEBUG:     - m_gs: 0x2b
DEBUG:     - m_ldtr: 0x0
DEBUG:     - m_tr: 0x40
DEBUG: 
DEBUG: segment base:
DEBUG:     - es_base(): 0x0
DEBUG:     - cs_base(): 0x0
DEBUG:     - ss_base(): 0x0
DEBUG:     - ds_base(): 0x0
DEBUG:     - fs_base(): 
----------------------------------------
- Standard Exception Caught            -
----------------------------------------
vector

So, it basically doesn't like something about FS