DVM fails to start after whonix update/dist-upgrade

eujan1 · October 29, 2016, 12:32am

System: Qubes 3.1 and Whonix 13.

Immediately after updating the whonix ws and gw templates today, whonix ws dvm failed to start. The DVM was created using the instructions at:

“Using Whonix-Workstation as a DisposableVM (DispVM)”

The DVM usually works perfectly although this is the second time this has happened and had to recreate it.
Originally assumed a one time only problem but apparently not the case.

Any idea as to what the cause might be? Can provide logs if they might help.

Possible relation to a problem after whonix updates to a normal whonix AppVM recently?

See:
“Whonix AppVM fails to start after update”

entr0py · October 29, 2016, 3:27am

Is the affected DVM based on one of the Whonix appVMs that fail to start? What makes you think the issues are related?

I haven’t experienced either of the issues you describe with a similar system.

Can you run some hardware diagnostics outside of Qubes? RAM, HDD…

Also, what is your graphics adapter?

eujan1 · October 31, 2016, 10:18pm

The DVM is (was) based on the Whonix-ws Template VM.[quote=“entr0py, post:2, topic:3109”]
What makes you think the issues are related?
[/quote]
Only the fact that it started immediatetly after a whonix-ws and gw update.

Extended RAM and HDD diagnostics show passed with 0 errors.
Checked motherboard, PCI, etc., while at it, all passed.[quote=“entr0py, post:2, topic:3109”]
Also, what is your graphics adapter?
[/quote]
Intel Broadwell-U Intigrated Graphics

Perhaps this is a KDE related problem?
Am getting 2 KDE popup errors on startup;

KDE crash handler

“Executable: kded4 PID: **** Signal; Bus error (7) Time; ************”

“Executable: kmix PID: **** Signal; Bus error (7) Time; ************”

Also, another KDE crash handler error if I try to open the application editor.

“Unable to contact khotkeys. Your changes are saved, but they could not be activated.”

“Executable: knotify4 PID: **** Signal; Bus error (7) Time; ************”

The editor then opens but the dvm entry is gone.

eujan1 · October 31, 2016, 10:36pm

Sorry, didn’t answer this question adequately.

The Whonix appVM that fails to start is based on the whonix-ws template as are other whonix appVMs that continue to function correctly.
The DVM is (was) based on the same Whonix-ws template.

Something else interesting that I just noticed is that in Qubes VM manager for the failed DVM, VM settings/advanced shows:

Initial memory:3412MB
Max memory: 4000MB
VCPUs no. 1

I did not change any defaults when the DVM was created.

entr0py · November 1, 2016, 4:20am

I’m sorry, I can’t be of much help here. The only issues I’ve had with KDE involved openGL compositing with an ATI card, but Intel Graphics seem to be well supported. I see you’re getting a bus error for kmix. I also had an issue with multiple audio adapters crashing the audio subsystem.

Wow, that is a huge chunk of RAM to be auto-assigned to the DVM… Mine is ~800 MB. I have no idea if that is the cause or the symptom of your issues though.

Patrick · November 1, 2016, 1:59pm

To debug this (or a VM that does not start generally…). (Do we have more generic debugging instructions on this @adw?)

Set QVMM to show internal and inactive VMs. (Qubes VM Manager -> View)

Then right after starting the VM, open a console.

sudo xl console name-vm-dispvm

(Such as Disp1 or so.)

See if there are any errors. Login as root. See if any systemd services failed loading.

sudo systemctl list-units --failed

eujan1 · November 2, 2016, 12:25am

Some strangeness happening…

The whonix-DVM would not start at all initially, then went to staying at the yellow phase if started directly from Qubes VM manager. Now it continues to the green phase if started from Qubes VM manager.
Regardless, I’m unable to start an app in the DVM, ie tor browser, which indicates " save file creation failed" and am unable to start a konsole session in the dvm.

Forgive my ignorance, but did you mean I should start a DOM0 konsole after starting the dvm @Patrick?

I tried that too and if so, the only error I can see with the above commands is an app-armor failure.

There are available logs from Qubes VM Manager if that might help?

Perhaps I should delete and recreate the DVM again but will wait for your answer in case you have other questions or would like to try some more debugging.

Patrick · November 2, 2016, 12:29am

Yes.

No failed systemd units?

adw · November 2, 2016, 4:03am

I’m afraid not (at least not that I’m aware of).

eujan1 · November 2, 2016, 11:46pm

Checked again to verify. No systemd units show failed.

Patrick · November 3, 2016, 4:04pm

Your help in debugging this is appreciated!

Please post the output of qvm-ls (run in dom0 terminal) after failed
start of dvm.

Please also check contents of files in dom0 /var/log/qubes containing
whonix-ws-dvm. To get a list of interesting log files you can use find /var/log/qubes | grep whonix-ws-dvm.

Same for /var/log/qubes folder in whonix-ws-dvm. (After you login using
xl console.)

eujan1 · November 4, 2016, 10:54pm

Thanks. Just a user here trying to muddle through.
The work you are doing is very much appreciated!

@dom0 ----------------------------- name | on | state | updbl | type | ----------------------------- {dom0} | * | Running | Yes | Admin | {sys-net} | * | Running | {sys-firewall} | * | Running | {sys-whonix} | * | Running | [debian-8] | | Halted | [whonix-ws] | | Halted | whonix-ws-dvm | * | Running | anon-whonix-bad | | Halted | whonix | | Halted | anon-whonix | | Halted | anon-whonix-clone-1 | * | Running | [whonix-gw] | | Halted | =>[fedora-23] | untrusted | * | Halted | personal | * | Halted | banking | | Halted | work-New | | Halted | Vault | | Halted | fedora-23-dvm | | Halted | ~]$ qvm-ls
-----±—±--------±------±------±------------------------±-------------±-------+
template | netvm | label |
-----±—±--------±------±------±------------------------±-------------±-------+
None | n/a | black |
| Net | fedora-23 | n/a | red |
| Proxy | fedora-23 | sys-net | green |
| Proxy | whonix-gw | sys-firewall | black |
Yes | Tpl | n/a | *sys-whonix | black |
Yes | Tpl | n/a | *sys-whonix | black |
| | whonix-ws | *sys-whonix | gray |
| | whonix-ws | sys-whonix | red |
| | whonix-ws | sys-whonix | orange |
| | whonix-ws | *sys-whonix | red |
| | whonix-ws | *sys-whonix | red |
Yes | Tpl | n/a | sys-whonix | black |
| Halted | Yes | Tpl | n/a | *sys-whonix | black |
| | fedora-23 | sys-firewall | red |
| | fedora-23 | - | yellow |
| | fedora-23 | sys-firewall | green |
| | fedora-23 | *sys-whonix | blue |
| | fedora-23 | - | black |
| | fedora-23 | sys-firewall | gray |

/var/log/qubes/guid.whonix-ws-dvm.log
Icon size: 128x128
invalid PMinSize for 0x3c00005 (0/0)
invalid PMaxSize for 0x3c00005 (0/0)
invalid PResizeInc for 0x3c00005 (0/0)
invalid PBaseSize for 0x3c00005 (0/0)
invalid PMinSize for 0x3c00006 (0/0)
invalid PMaxSize for 0x3c00006 (0/0)
invalid PResizeInc for 0x3c00006 (0/0)
invalid PBaseSize for 0x3c00006 (0/0)
invalid PMaxSize for 0x3c00006 (0/0)
invalid PResizeInc for 0x3c00006 (0/0)
libvchan_is_eof
Icon size: 128x128

/var/log/qubes/guid.whonix-ws-dvm.log.old
Icon size: 128x128
invalid PMinSize for 0x4400005 (0/0)
invalid PMaxSize for 0x4400005 (0/0)
invalid PResizeInc for 0x4400005 (0/0)
invalid PBaseSize for 0x4400005 (0/0)
invalid PMinSize for 0x4400006 (0/0)
invalid PMaxSize for 0x4400006 (0/0)
invalid PResizeInc for 0x4400006 (0/0)
invalid PBaseSize for 0x4400006 (0/0)
invalid PMaxSize for 0x4400006 (0/0)
invalid PResizeInc for 0x4400006 (0/0)
libvchan_is_eof
Icon size: 128x128
domain dead
Failed to connect to gui-agent

/var/log/qubes/pacat.whonix-ws-dvm.log
0 bytes

/var/log/qubes/pacat.whonix-ws-dvm.log.old
0 bytes

/var/log/qubes/qrexec.whonix-ws-dvm.log
Rpc allowed: whonix-ws-dvm dom0 qubes.WindowIconUpdater
Rpc allowed: whonix-ws-dvm dom0 qubes.NotifyUpdates
eintr
libvchan_is_eof

/var/log/qubes/qubesdb.whonix-ws-dvm.log
vchan closed
reconnecting
vchan closed

This produced no info.
@dom0 ~]$ sudo xl console whonix-ws-dvm

user@host:~$ find /var/log/qubes | grep whonix-ws-dvm
user@host:~$

Patrick · November 5, 2016, 12:10am

Can you make head or tail of this report or could use any further debug output? @marmarek

eujan1 · November 8, 2016, 12:15am

If more debug info is desired, please let me know. If not, will delete the affected dvm and recreate.