Possibility to detect suspend events in VM?

[Phil Knüfer]

Hi,

when accessing an SMB share in a Qubes AppVM after the system has been
suspended for some time, I experience a serious lag (up to about a
minute or so for each share that is mounted from the same SMB server).
This seems to be due to the fact that the server has already timed out
the SMB session, while the client (the AppVM) is still trying to resume
it and therefore runs in TCP timeouts.

For bare-metal Linux systems, a possible solution is to unmount all SMB
shares before the system goes into suspend (e.g., via a script in
/usr/lib/systemd/system-sleep or via pm-utils).

I tried this approach in Qubes but it seems that the AppVMs do not know
about a suspend event. Is there a way to trigger scripts on suspend in
Qubes AppVMs or do I need to coordinate the SMB unmount from dom0 (it
should be possible to trigger a script there that interacts with VMs via
qvm-run or similar)?


Regards,
Phil

[unman]

I think that the dom0 route is the way to go - it's what I use myself.
If you did find a way in the qube of detecting such events, I'd be
interested.

[Phil Knüfer]

Am 12.07.20 um 14:37 schrieb unman:

I just had the time to look into this. Analyzing journalctl in an AppVM
I saw the line:

Jul 22 18:45:04 work qrexec-agent[6750]: executed root:QUBESRPC
qubes.SuspendPreAll dom0 pid 6753

A quick web search yielded this Github issue:
https://github.com/QubesOS/qubes-issues/issues/1663


TL;DR: I haven't tested it yet but it seems placing the PRE- and
POST-suspend files in /etc/qubes/suspend-pre.d/ and
/etc/qubes/suspend-post.d/ should do the trick.


Regards,
Phil