-
-
Notifications
You must be signed in to change notification settings - Fork 48
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
sys-net occasionally dies on resume from suspend #4658
Comments
Posted bounty ($100): https://www.bountysource.com/issues/67959138-sys-net-dies-on-resume-from-suspend |
@andrewdavidwong @brycepg I feel this must be hardware specific, since I dont see it on any of the machines that I still have access to running 3.2.1. |
Laptop model: Lenovo Thinkpad T530 EDIT: Strangely I cannot repro right now with a quick suspend. Will try out with longer suspend times. |
I should have clarified that this only happens occasionally, so it would be difficult to reproduce without suspending/resuming many times. I'll update the issue to reflect this. Lenovo Thinkpad T450s:
|
same issue, thinkpad x220, qubes 4.0 Maybe a script that shut down and restart all net-vm could be a temporary fix ? ( its pretty long and painful to do it manually ) |
This happened to me recently after re-doing some templates and VMs. What I discovered was that my wifi module suspend settings were gone.... I forgot to re-add them. I agree with unman this is a hardware-specific issue. OTOH it would be nice if Qubes had some way of automatically populating this module information. Here is what I use in /rw/config/suspend-module-blacklist for an Intel "Ultimate-N" card:
|
In my case,
The result in my case is the following output in sys-net: In the case of I hope that this is relevant for this issue. If not please indicate how I can proceed. For the record, essentially the machine (DELL XPS 13 9360) is running R4.0. |
I'm experiencing this issue with a Librem 13 laptop and my easiest solution is to kill the |
Myself and several peers all have librem15 and librem13 devices and sys-net dies often on resume exactly as @quantumpacket describes. This is a serious PITA we would love help with. |
As per the discussion on qubes-users ("[qubes-users] sys-net keeps dying") here is the dmesg from a netvm (formerly fedora-29, upgraded to fedora-30) as it begins to exhibit the described behaviour:
|
Creating a new sys-net does not appear to have fixed the issue, crashes still occur. |
I also face this issue sometimes but also with sys-usb. I also have a T450. I will post the the ouput of
The next time it occurs. |
@andrewdavidwong Could you claim the bounty I made please? Apparently BountySource will take the bounty after 2 years which is 2 months away, and I want it to go to a Qubes member. I've pretty much never had this issue after getting a new (used) T530 which supports HVM so I could install Qubes 4.0 |
Thanks for letting us know, @brycepg. I think the bounty should be turned into a donation to the Qubes OS Project, if possible. @mfc, @MiCh, do you know how to do that? |
I also have not experienced this bug in a very long time, even on the same hardware as when I filed this report. It looks like the last report was from @w1k1n9cc on May 24. @w1k1n9cc, are you still experiencing this? |
My Qubes-PC is not very active at the moment. I will try it until tuesday. Maybe I have some sparse time at the weekend. |
@andrewdavidwong I think the reason everyone came here is that this issue was still open, and the other was closed. Now they are both closed. But the issue is clearly not fixed. |
@DemiMarie, you're the one who asked whether this is a duplicate of #4042. Do you no longer have reason to think it is? If so, what are those reasons? And please don't say "because the other one is closed." That would be a reason to reopen the other issue, not this one! |
Whoops! |
@andrewdavidwong #4042 affects sys-usb whie this one affects sys-net. |
I tried disabling this service via |
Disabling systemd-udevd will break a lot of stuff. |
On 5/11/22 20:07, Demi Marie Obenour wrote:
>> Try disabling the `/usr/lib/systemd/system/systemd-udevd.service` watchdog in your `sys-net` template and please report back if it helps.
>
> I tried disabling this service via `systemd stop systemd-udevd && systemd disable systemd-udevd` and it didn't fix the issue. Disabling the service via Qubes VM settings (adding an entry for `systemd-udevd` and unchecking the box) also did not help.
Disabling systemd-udevd will break a lot of stuff.
Exactly.
I was talking about the watchdog back then, not the entire service. I.e. remove or comment out the `WatchdogSec` line in that file. I was suspecting that a udevd restart caused by the watchdog (which is triggered by clock issues caused by the resume from suspend) would cause havoc.
|
On Wed, May 11, 2022 at 11:28:33AM -0700, 3hhh wrote:
I was talking about the watchdog back then, not the entire service. I.e. remove or comment out the `WatchdogSec` line in that file. I was suspecting that a udevd restart caused by the watchdog (which is triggered by clock issues caused by the resume from suspend) would cause havoc.
There is no WatchdogSec line. Here is my systemd-udevd.service file:
```
[Unit]
Description=Rule-based Manager for Device Events and Files
Documentation=man:systemd-udevd.service(8) man:udev(7)
DefaultDependencies=no
After=systemd-sysusers.service systemd-hwdb-update.service
Before=sysinit.target
ConditionPathIsReadWrite=/sys
[Service]
DeviceAllow=block-* rwm
DeviceAllow=char-* rwm
Type=notify
# Note that udev will reset the value internally for its workers
OOMScoreAdjust=-1000
Sockets=systemd-udevd-control.socket systemd-udevd-kernel.socket
Restart=always
RestartSec=0
ExecStart=/usr/lib/systemd/systemd-udevd
ExecReload=udevadm control --reload --timeout 0
KillMode=mixed
TasksMax=infinity
PrivateMounts=yes
ProtectClock=yes
ProtectHostname=yes
MemoryDenyWriteExecute=yes
RestrictAddressFamilies=AF_UNIX AF_NETLINK AF_INET AF_INET6
RestrictRealtime=yes
RestrictSUIDSGID=yes
***@***.*** @module @raw-io
SystemCallErrorNumber=EPERM
SystemCallArchitectures=native
LockPersonality=yes
IPAddressDeny=any
```
|
Well, then it's not that on your system.
(Debian has the watchdog btw, Fedora apparently not.)
|
Still experiencing this problem. In my case, it ONLY affects WiFi, Ethernet connections continue normally. See also #5508 |
I haven't had this happen to be for a while. And I haven't had it happen to me at all since upgading to 4.1 (Lenovo T530). My sys-net is on |
Fedora 35 is EOL, FYI |
This issue is being closed because:
If anyone believes that this issue should be reopened, please leave a comment saying so. |
Qubes OS version:
R3.2
Affected component(s):
sys-net
Steps to reproduce the behavior:
Expected behavior:
sys-net
stays on and continues to provide network access normally.Actual behavior:
Occasionally:
sys-net
looks like it's trying to connect for a second, then the entiresys-net
just dies (shows powered off in Qubes Manager).sys-firewall
and AppVMs usingsys-firewall
for network access are still running normally, but of course they don't have network access.sys-net
does not restore network access to these other AppVMs.sys-net
. (Maybe restarting justsys-net
andsys-firewall
would be enough, but in practice it's easier for me just to shut them all down and restart them all.)General notes:
This has been going on for at least a few months. When reading #4657, I saw that it mentioned this problem with
sys-net
in passing. I thought we already had an issue on this (see below) but couldn't find one, so I'm filing this now.Related issues:
I could have sworn we already had an issue about this, but after searching, I can't find one. These all look different:
#2964 was about losing network access when
sys-net
stays on#3008/#3030 was about failing to connect to a network after resume when
sys-net
stays on#3151 was about NetworkManager not running in
sys-net
after resume whensys-net
stays on#3738 was about the entire computer not resuming correctly from suspend.
Ah, maybe I was thinking of #4042, which is a similar report about
sys-usb
.The text was updated successfully, but these errors were encountered: