FS#56785 - [irqbalance] do not ship ugly udev rule new in 1.3.0-1

Attached to Project: Arch Linux
Opened by Jason William Walton (jasonww) - Wednesday, 20 December 2017, 06:48 GMT
Last edited by David Runge (dvzrv) - Thursday, 16 May 2019, 21:10 GMT
Task Type General Gripe
Category Packages: Extra
Status Closed
Assigned To Dan McGee (toofishes)
Anatol Pomozov (anatolik)
Architecture All
Severity Low
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 1
Private No

Details

irqbalance in testing is shipping a new sample udev rule that is terrible:

https://github.com/Irqbalance/irqbalance/commit/10bb3d63d41e2b47fe24aa8d4a5dca4cd0478b62#diff-0a5da898d9943cf9cca2eada25e6a214

ACTION=="add|remove", RUN+="/usr/bin/killall -SIGHUP irqbalance"

Obviously this is horrible and not required for irqbalance to work. If people need very fast responses to irq changes they can can add it manually. Alternative irqbalance should handle all of this without resorting to this hugely ugly garbage.
This task depends upon

Closed by  David Runge (dvzrv)
Thursday, 16 May 2019, 21:10 GMT
Reason for closing:  Not a bug
Additional comments about closing:  The udev rule is not included in the package.
Comment by glavin lewis (Glavin) - Wednesday, 20 December 2017, 14:06 GMT
Ya I second that, journal log gets filled with "systemd-udevd[193]: Process '/usr/bin/killall -SIGHUP irqbalance' failed with exit code 1" and at the end irqbalance doesn't even start. You have to start it manually.

btw it's in extra's now.
Comment by Doug Newgard (Scimmia) - Wednesday, 20 December 2017, 15:24 GMT
Other than the misappropriation of SIGHUP, how is this "ugly" "terrible" "horrible" and "garbage"?
Comment by Anatol Pomozov (anatolik) - Wednesday, 20 December 2017, 17:49 GMT
Is it really that useless? Or there is a way to improve this rule?

This udev rule comes from upstream https://github.com/Irqbalance/irqbalance/blob/master/misc/90-irqbalance.rules And the best of all if it gets fixed (or removed) upstream, in this case all Linux distros get benefit of it. Could you please file an upstream issue and track it there?
Comment by Jason William Walton (jasonww) - Wednesday, 20 December 2017, 18:15 GMT
irqbalance was in the repos almost 10 years without this behavior and as far as iam concernced this looks very much like an optional rule. people who care about irqbalance not working for a minute or so when you hotplug a pcie card already know about interrupt management. obviously there are ways to make it less terrible, but iam not a udev person but usage of ATTRS{irq}=="*" or something like that would go a long way here. irqbalance needs to ship a helper binary or handle this in demon, calling killall from udev is unacceptable.
Comment by Doug Newgard (Scimmia) - Wednesday, 20 December 2017, 18:23 GMT
I think you just don't know what killall -SIGHUP does.
Comment by Jason William Walton (jasonww) - Wednesday, 20 December 2017, 18:25 GMT
yeah maybe you think calling a binary a billion times on boot without any error checking if irqbalance runs that scans the entire process tree for a process that doesnt even exist is cool then maybe go work for microsoft.
Comment by Doug Newgard (Scimmia) - Wednesday, 20 December 2017, 18:53 GMT
https://en.wikipedia.org/wiki/Moving_the_goalposts
Comment by Francois (francoism90) - Wednesday, 05 September 2018, 08:21 GMT
Can this be closed? We are at 1.3.0-2 now. And any update on the new release? :)

Loading...