Greetings, it has been a while since I have posted, and I wanted to give an update and ask a related question to see if I can figure out this last remaining network problem I have had since I started using Qubes OS.
I have used Qubes for almost two years now after a very painful and unproductive attempt to get a Windows 10 guest working on Xen via the manual method.
This last real issue it has is more annoying than an actual problem, as it only happens after a week or so, and it is easily fixed with a simple process.
The gist of the problem is that Win 10 guests lose network access after 10 days of uptime. In the past, it seemed like it was random, but after about a dozen notations of adapter uptime when it fails, it seems to almost always be 10 days and then the network just stops. Connected apps immediately error out with network errors (Putty aborts instantly, other apps may take a few seconds to a minute to detect the loss of network connectivity).
Qubes itself does not lose network access, so I know it isn’t the machine hardware, Qubes/Xen/Linux, or the rest of my network setup causing the problem.
Simply restarting the Win 10 Qube does not fix it, however. I have to go into Device Mangler and Uninstall BOTH the Xen PV Network driver (under Network Devices), AND the Xen PV Network Class driver (under System Devices) and then restart twice, letting Windows re-install both drivers. Once in a while it takes removing it a second or third time to take, but usually just once and the network comes back up on the first restart (the second restart is just to satisfy the Xen PV driver install restart request, just to be sure everything is copacetic in the Windows driver stack).
After many moons, I have pondered why it might always be crashing at nearly exactly 10 days of adapter/system uptime, and the only thing that comes to mind is something like an IP lease from Qubes/Xen to the guest that expires, but the only way to renew it fully is to play uninstall-the-drivers game. I have checked the Qubes doc and it says that DHCP is not used to assign IPs to the guests, so unless there’s some kind of emulation of IP lease management, it isn’t that.
I suppose it could be something endemic to the Xen PV network drivers, but why 10 days? Is it overflowing a time accumulator and crashing or something? There is nothing in the event logs to indicate a driver failure or other event, and the nature of the failure seems to be reported immediately to apps when the time limit elapses. Every app is working fine and then vwoop! No net. The icon in the tray immediately changes from the connected to internet plug symbol to the not-connected world globe symbol, almost as if I pulled the net cable out of the back of the computer.
Again, it’s not debilitating, just slightly annoying. I’ve sort of planned on the day I know it is set to expire to not be doing anything super important at the time I expect it to fail, but I would like to figure out what in the world is causing it and see if I can fix it.
Win 10 on Qubes has been my daily driver for two years now, and outside of some initial pains/freezes and performance issues (I still have some minor ones), it’s been very stable and reliable. I’m happy with it, and I have high hopes for going to 4.2/4.3 once the QWT situation is sorted.
Thanks for your time, attention, and any thoughts you might have on the subject!