Windows 10 VM reboots when starting software

jstorgaard

@olivierlambert said in Windows 10 VM reboots when starting software:

If it's MSR related, you should see xl dmesg output when the VM crashes.

This is output from xl dmesg when machine crashes:

(XEN) [57920.851786] d10v1 Unknown Host LBR MSRs
(XEN) [57920.851789] domain_crash called from vmx.c#vmx_msr_write_intercept+0x4c2/0x510
(XEN) [57920.851790] Domain 10 (vcpu#1) crashed on cpu#8:
(XEN) [57920.851791] ----[ Xen-4.13.0-8.6.1.xcpng8.1  x86_64  debug=n   Not tainted ]----
(XEN) [57920.851792] CPU:    8
(XEN) [57920.851793] RIP:    0010:[<fffff8014a7de482>]
(XEN) [57920.851794] RFLAGS: 0000000000040002   CONTEXT: hvm guest (d10v1)
(XEN) [57920.851795] rax: 0000000000000001   rbx: 0000000000000000   rcx: 00000000000001d9
(XEN) [57920.851796] rdx: 0000000000000000   rsi: 0000000000000000   rdi: 00000000000000f0
(XEN) [57920.851797] rbp: fffff68839bd7b00   rsp: fffff68839bd7a78   r8:  0000000000000001
(XEN) [57920.851798] r9:  fffff68839bd75f8   r10: 0000000000000000   r11: ffffd50f3594a080
(XEN) [57920.851799] r12: 0000000002a4b528   r13: 00007ff907ec4900   r14: 00000000000000f0
(XEN) [57920.851799] r15: 0000000000000000   cr0: 0000000080050033   cr4: 0000000000370678
(XEN) [57920.851800] cr3: 0000000061fa4002   cr2: 000000000027425a
(XEN) [57920.851801] fsb: 0000000000000000   gsb: ffffb980e01d1000   gss: ffffb980e01d1000
(XEN) [57920.851802] ds: 002b   es: 002b   fs: 0053   gs: 002b   ss: 0018   cs: 0010

jstorgaard

@olivierlambert said in Windows 10 VM reboots when starting software:

Also, please share the BSOD screen if possible

Unfortunately no BSOD data available. No guest logs or entries in EventLog. There are never any BSOD shown in the client.
It just plain shuts off or reboots from XCP-ng.

olivierlambert

You might have a Windows setting to prevent reboot in crash (or also with xe you can set the VM to not restart in case of crash).

Thanks for the output

olivierlambert

Can you do a copy/paste (eg in hastebin.com) the full output of xl dmesg from your dom0?

jstorgaard

@olivierlambert said in Windows 10 VM reboots when starting software:

Can you do a copy/paste (eg in hastebin.com) the full output of xl dmesg from your dom0?

Here is the complete dmesg:
https://hastebin.com/ruvuqokiba.sql

jstorgaard

@olivierlambert said in Windows 10 VM reboots when starting software:

You might have a Windows setting to prevent reboot in crash (or also with xe you can set the VM to not restart in case of crash).

Thanks for the output

I removed the mark that makes Windows boot in case of a crash, but same result. It just acts like it's a hypervisor crash and not a Windows crash.
How can I change the setting in xe? Can't seem to find that option in XO.

olivierlambert

Yeah in xe, it's actions_after_crash that you should set to preserve but in your case, it might be not even relevant.

Let me take a look at the dmesg (and showing it to some Xen devs)

olivierlambert

Okay @jstorgaard I have a good news. It's likely related to your CPU.

Family 6 Model 165 Stepping 3 in your case, is a Comet Lake CPU.

And Xen version you have don't have yet the right code for some calls asked by the VM, trigger a crash (Xen doesn't know which MSRs to give it).

This was patched in Xen relatively recently (few months ago), ask you can see here: https://xenbits.xen.org/gitweb/?p=xen.git;a=commitdiff;h=1fe406685cb19e9544681c6243e7d376deb0297e

This patch is relatively trivial, so in theory we could backport it and build an updated Xen version. Then put the package in testing repo so you can test it and confirm it solves the issue.

I don't know if @r1 can build it for us, but we'll have to wait next week for @stormi to put the package in testing anyway

edit: issue created: https://github.com/xcp-ng/xcp/issues/434

olivierlambert created this issue in xcp-ng/xcp

closed Backport Xen patch to support Ice Lake and Comet Lake CPUs #434

jstorgaard

@olivierlambert Thanks a lot, that's good news!
I'm very much up for testing so I'll keep an eye on the issue you created

olivierlambert

A big thanks to @andyhhp for his priceless help on this

r1

@olivierlambert said in Windows 10 VM reboots when starting software:

I don't know if @r1 can build it for us, but we'll have to wait next week for @stormi to put the package in testing anyway

The patch needed a rework for fitment. I'd create a PR for modified patch and then it can go in testing repo.

stormi

@jstorgaard Updated packages are now available for you to test: https://github.com/xcp-ng/xcp/issues/434#issuecomment-688786905

olivierlambert created this issue in xcp-ng/xcp