-
Notifications
You must be signed in to change notification settings - Fork 5k
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
Instabilities/crashes due to kernel update 5.4.51 #3756
Comments
Probably same context as #3757 |
I also reported an issue related to the kernel update breaking a dkms package: RPi-Distro/repo#188 I was expecting RaspberryPi OS Buster to follow what Debian Buster does (they stick with 4.19 kernels). It seems diverting from them and update to 5.4 kernels wasn't the best decision, and it was done without proper testing, otherwise the dkms failures would have been detected. |
We need to move to 5.4 for many reasons, the newer Mesa stuff which is needed for our graphics stack is one major one, and that cannot be backported in any feasible way. So the move to 5.4 was unavoidable. This 5.4 branch has been available for testing for 4 months in github, and in rpi-update for some weeks. So anyone could have tried it out and reported any issues. We do rely on third parties for testing as we are a very small team with lots of other work to do. Unfortunately, testing doesn't catch everything. |
Sorry @JamesH65 , I wasn't aware of that. So, now that a kernel 5.4 is necessary and we know is breaking other packages, would it be possible to get those packages fixed? For instance, for the |
We are working right now on fixing the issues that have surfaced. There are some high priority ones to sort out before this one I suspect. We would also take PR's if others can fix the issues before we get round to them. |
@vincent-olivert-riera thanks for mentioning the upstream |
You won't be able to install |
Describe the bug
On an originally stable system (2 voluntary restarts in 20 months).
Unable to compile kernel modules vs. kernel headers:
Describe the bug
I use the xtables-addons-common package (v3.2-1 from the repository) to complete my server's iptables firewall.
This important package requires the current kernel headers. However, multiple errors appear when compiling it
using dkms (xtables-addons-dkms package).
To reproduce
In the make.log file I distinguish 2 types of errors:
Kernel independent errors (4.19.75 and 5.4.51) in the precompiled binaries in the
/srcripts/
directory .To reproduce
If manual execution is attempted:
I thought at first I'd have to recompile the kernel. In fact It was a waste of time since they are actually binaries that don't match the platform of the Raspberry Pi.
Tests performed in /usr/src/linux-headers-*/
test on binary from the raspberrypi-kernel-headers package:
test on compiled cross binary:
Expected behaviour
test binary compiled on site:
Actual behaviour
make scripts
command though useful because it could avoid getting lost in useless compilations.(see https://www.raspberrypi.org/documentation/linux/kernel/building.md).
Errors specific to version 5.4.51:
Actual behaviour
The xtables-addons v3.2 module sources are incompatible with 5.x kernels.
According to https://sourceforge.net/p/xtables-addons/xtables-addons/commit_browser, the commit that provides this support
was integrated in version 3.6 in November 2019.
I decided to install via dpkg the .debs from Debian Bullseye repositories (the package doesn't seem to have been integrated into Buster?): https://packages.debian.org/bullseye/xtables-addons-common.
Dkms agrees to compile but I do not guarantee the results.
Expected behaviour
If kernel version 5.4.x is to be deployed in repositories then the xtables packages should be updated.
5.4.51 Kernel related instabilities
Describe the bug
The system became very unstable after the reboot:
very frequent crashes, sometimes even without having time to do ssh login
maximum uptime of about 40 minutes
bad evaluation of the CPU loads obtained with the uptime command (~4 for a CPU at rest)
crazy processes; Example of Gunicorn log:
[2020-07-25 11:11:39 +0200] [736] [INFO] Listening at: unix:/run/umatrix.sock (736)
[2020-07-25 11:11:39 +0200] [736] [INFO] Using worker: sync
[2020-07-25 11:11:39 +0200] [1127] [INFO] Booting worker with pid: 1127
[2020-07-25 11:11:43 +0200] [1165] [INFO] Booting worker with pid: 1165
[2020-07-25 11:42:45 +0200] [740] [INFO] Starting gunicorn 19.9.0
[2020-07-25 11:42:46 +0200] [740] [INFO] Listening at: unix:/run/umatrix.sock (740)
[2020-07-25 11:42:46 +0200] [740] [INFO] Using worker: sync
[2020-07-25 11:42:46 +0200] [1124] [INFO] Booting worker with pid: 1124
[2020-07-25 11:43:34 +0200] [740] [INFO] Handling signal: term
[2020-07-25 11:43:34 +0200] [1124] [INFO] Worker exiting (pid: 1124)
[2020-07-25 11:43:34 +0200] [740] [INFO] Shutting down: Master
... and so on for hundreds of loops...
non-startable processes via systemd
process persistent in "S - interruptible sleep" in htop after a sigkill/sigterm
At first I thought there were regressions in the updated packages since the last reboot in October 2019.
To reproduce
Do not know.
Things tested:
No improvement was seen between each of these tests.
Below you will find the only kernel panic stacktraces I was able to retrieve;
I'm at your disposal for any other logs that I could bring in extra data.
Actual behaviour
(Thank you for keeping all versions of these packages, I don't know if there is a method to get them back with apt ?).
System
Linux bbox-NAS 4.19.75-v7+ #1270 SMP Tue Sep 24 18:45:11 BST 2019 armv7l GNU/Linux
console=tty1 root=PARTUUID=XXXXXXX-XX rootfstype=ext4 elevator=deadline fsck.repair=yes rootdelay rootwait quiet
Logs
Best regards!
The text was updated successfully, but these errors were encountered: