Hi, Currently, the Mageia kernel only has Cubic hardcoded as the only congestion control algo. I would like to request to add the others too such as illinois, yeAH, westwood, etc. It is unclear to me why your kernel does not have them included while other distro's kernel offer them all.
CC: (none) => neutrino8
> request to add the others too such as illinois, yeAH, westwood, etc > other distro's kernel offer them all How does one choose between them? Clearly this is for tmb to reply to.
CC: (none) => lewyssmith, tmb
you can change the CCA by issuing the following (example): echo "illinois" > /proc/sys/net/ipv4/tcp_congestion_control The problem is, you can't do that since the Mageia kernel has cubic hardcoded as the only available CCA. Adding support for the others is just a matter of selecting them during kernel config. They'll be built as modules.
But there should be a more human way of causing the example command you cite, which is very expert; something in MCC in our case? How do other distributions make the available choices known and selectable?
You can add in /etc/sysctl.conf which CCA you want, something like: net.ipv4.tcp_congestion_control = illinois I think this can be easily implemented in MCC. I'm currently on openSUSE which has YaST and while it has no option for such, I just use sysctl.conf to select the CCA I want. But first, the kernel has to offer the others too and not just cubic.
(In reply to Grozdan Nikolov from comment #0) > Hi, > > Currently, the Mageia kernel only has Cubic hardcoded as the only congestion > control algo. I would like to request to add the others too such as > illinois, yeAH, westwood, etc. > > It is unclear to me why your kernel does not have them included while other > distro's kernel offer them all. Basically because no-one have requested them / expressed the need so far :) config updates queued in svn for next kernel builds for Cauldron and Mga 8 Cubic will still be the default one, but you will atleast be able to change if/when needed (In reply to Lewis Smith from comment #3) > But there should be a more human way of causing the example command you > cite, which is very expert; something in MCC in our case? or not... theese are expert options that you actually need to know what they do before you start tuning them... not really something a "normal end-user" need to even think about...
They're no more expert than say choosing a file system to use or choosing an I/O scheduler to use for an SSD/disk.
well, they are more expert (or maybe I should call them advanced) in the sense that most end-users dont really hit tcp congestion in a way that they would notice and theese would solve since they are still at the mercy of whatever / however their isp treats their traffic... for filesystems many/most? endusers tend to go with the distro default for simplicity... disk i/o scheduling on the other hand can actually be noticeable for the enduser... but many endusers dont really care about that either...
There is now a kernel-5.15.60-1.mga8 in Core Updates Testing that has enabled support for switching algos
Thanks a lot! :) I consider this topic closed...
(In reply to Thomas Backlund from comment #8) > There is now a kernel-5.15.60-1.mga8 in Core Updates Testing that has > enabled support for switching algos Impressive work Thomas, and thanks a lot. Assigning to QA in consequence. @Grozdan When you can try the new kernel 5.15.60 (best, from Core/Updates_Testing; else when it gets released as an update), please do test the functionality you asked for in comment 0, supplemented by comment 2, and report back here. Our normal kernel testers are unlikely to be able to test the efficacity of the change.
Assignee: bugsquad => qa-bugs
Too fast... Please dont assign kernels bugs to qa ... I assing update requests when I consider them ready for wider tests, not for every bug that gets fixed...
Assignee: qa-bugs => kernel
After running "modprobe -v tcp_bic", changed from cubic to bic and back again. [root@x3 ~]# echo bic>/proc/sys/net/ipv4/tcp_congestion_control [root@x3 ~]# cat /proc/sys/net/ipv4/tcp_congestion_control bic [root@x3 ~]# echo cubic>/proc/sys/net/ipv4/tcp_congestion_control [root@x3 ~]# cat /proc/sys/net/ipv4/tcp_congestion_control cubic Will close this bug when kernel 5.15.60-$flavor-4.mga8 is pushed to updates.
CC: (none) => davidwhodgins
Depends on: (none) => 30750
Hi, It's been some time and a new kernel hasn't been released that offers these modules. I see there's already 5.15.61 in -testing but I like to get my updates from the official channel. Any ideas when we can expect it?
An update for this issue has been pushed to the Mageia Updates repository. https://advisories.mageia.org/MGASA-2022-0305.html
Status: NEW => RESOLVEDResolution: (none) => FIXED