CONFIG_TCP_CONG_BBR not set on, in fc32 kernel confg

[Joel Wirāmu Pauling]

Hi there;

I noticed that the kernel builds on the branched fedora32 kernels do not
have CONFIG_TCP_CONG_BBR set ; meaning no shiny.

--

root@kuriiti network-scripts]# sysctl
net.ipv4.tcp_available_congestion_control
net.ipv4.tcp_available_congestion_control = reno cubic
[root@kuriiti network-scripts]# uname -a
Linux kuriiti.example.com 5.6.0-0.rc5.git0.2.fc32.x86_64 #1 SMP Tue Mar 10
19:09:42 UTC 2020 x86_64 x86_64 x86_64 GNU/Linux
[root@kuriiti network-scripts]#

--

Apologies if this has been discussed previously, but this I believe is an
omission.

[Peter Robinson]

> I noticed that the kernel builds on the branched fedora32 kernels do not
> have CONFIG_TCP_CONG_BBR set ; meaning no shiny.

The kernel module is built:
$ modinfo tcp_bbr
filename:
/lib/modules/5.6.0-0.rc5.git0.2.fc32.x86_64/kernel/net/ipv4/tcp_bbr.ko.xz
description:    TCP BBR (Bottleneck Bandwidth and RTT)
> root@kuriiti network-scripts]# sysctl
> net.ipv4.tcp_available_congestion_control
> net.ipv4.tcp_available_congestion_control = reno cubic

If you manually load the module it's there:

# sysctl -a | grep tcp_available_congestion_control
net.ipv4.tcp_available_congestion_control = reno cubic
# modprobe tcp_bbr
# sysctl -a | grep tcp_available_congestion_control
net.ipv4.tcp_available_congestion_control = reno cubic bbr

I'm not sure the "official" way to use different congestion control
algorithms but I'm guessing it's something you can use NetworkManager
to specify and autoload them.

Peter

[Joel Wirāmu Pauling]

Thanks ; I guess a decision was made to build it as a module vs included in
baseline at some point. I hadn't even thought of checking modules.

On Tue, 17 Mar 2020 at 22:56, Peter Robinson  wrote:
> > I noticed that the kernel builds on the branched fedora32 kernels do not
> > have CONFIG_TCP_CONG_BBR set ; meaning no shiny.
>
> The kernel module is built:
> $ modinfo tcp_bbr
> filename:
> /lib/modules/5.6.0-0.rc5.git0.2.fc32.x86_64/kernel/net/ipv4/tcp_bbr.ko.xz
> description:    TCP BBR (Bottleneck Bandwidth and RTT)
> > root@kuriiti network-scripts]# sysctl
> > net.ipv4.tcp_available_congestion_control
> > net.ipv4.tcp_available_congestion_control = reno cubic
>
> If you manually load the module it's there:
>
> # sysctl -a | grep tcp_available_congestion_control
> net.ipv4.tcp_available_congestion_control = reno cubic
> # modprobe tcp_bbr
> # sysctl -a | grep tcp_available_congestion_control
> net.ipv4.tcp_available_congestion_control = reno cubic bbr
>
> I'm not sure the "official" way to use different congestion control
> algorithms but I'm guessing it's something you can use NetworkManager
> to specify and autoload them.
>
> Peter

[Justin Forbes]

On Tue, Mar 17, 2020 at 6:31 AM Joel Wirāmu Pauling  wrote:
> Thanks ; I guess a decision was made to build it as a module vs included in
> baseline at some point. I hadn't even thought of checking modules.
>
> This is not something that has changed, git history

on  configs/fedora/generic/CONFIG_TCP_CONG_BBR show last change was the
general config reworking in 2017.  Even this did not change the way it was
set, only how our config files are generated.  Curious if something else
has changed in how userspace tries to load the modules.

Justin

On Tue, 17 Mar 2020 at 22:56, Peter Robinson  wrote:
> > > I noticed that the kernel builds on the branched fedora32 kernels do
>
> not
> > > have CONFIG_TCP_CONG_BBR set ; meaning no shiny.
> >
> > The kernel module is built:
> > $ modinfo tcp_bbr
> > filename:
> > /lib/modules/5.6.0-0.rc5.git0.2.fc32.x86_64/kernel/net/ipv4/tcp_bbr.ko.xz
> > description:    TCP BBR (Bottleneck Bandwidth and RTT)
> > > root@kuriiti network-scripts]# sysctl
> > > net.ipv4.tcp_available_congestion_control
> > > net.ipv4.tcp_available_congestion_control = reno cubic
> >
> > If you manually load the module it's there:
> >
> > # sysctl -a | grep tcp_available_congestion_control
> > net.ipv4.tcp_available_congestion_control = reno cubic
> > # modprobe tcp_bbr
> > # sysctl -a | grep tcp_available_congestion_control
> > net.ipv4.tcp_available_congestion_control = reno cubic bbr
> >
> > I'm not sure the "official" way to use different congestion control
> > algorithms but I'm guessing it's something you can use NetworkManager
> > to specify and autoload them.
> >
> > Peter
>
> --
> Joel Wirāmu Pauling
> Senior Solutions Architect
> Mobile: [phone]
> Email: [e-mail] 
> _______________________________________________
> kernel mailing list -- [e-mail]
> To unsubscribe send an email to [e-mail]
> Fedora Code of Conduct:
> https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> List Archives:
> https://lists.fedoraproject.org/archives/list/[e-mail]

[Joel Wirāmu Pauling]

At one point there were some requirements on needing fq as the default
scheduler algorithm; IIRC when BBR code got merged in mainline in 4.14
that requirement was dropped. I haven't been playing in the network space
much in the last 2 years so stopped paying too much attention.

Generally in the past load workflow is to have the scheduler set as a
sysctl config file with a high(low) order preference so it gets applied
early in boot. I haven't noticed any issues with this and modprobe bringing
the module up on the two fc32 nodes I am using. But I did have issues back
when I was doing extensive testing in 2017 of the BBR with self compiled
kernels to do with ramdisk inclusion/rescue modes etc - which is one of the
reasons I started to always compile it in.

-Joel

On Wed, 18 Mar 2020 at 03:38, Justin Forbes  wrote:
> On Tue, Mar 17, 2020 at 6:31 AM Joel Wirāmu Pauling 
> wrote:
> > Thanks ; I guess a decision was made to build it as a module vs included
> > in
> > baseline at some point. I hadn't even thought of checking modules.
> >
> > This is not something that has changed, git history
>
> on  configs/fedora/generic/CONFIG_TCP_CONG_BBR show last change was the
> general config reworking in 2017.  Even this did not change the way it was
> set, only how our config files are generated.  Curious if something else
> has changed in how userspace tries to load the modules.
>
> Justin
>
> On Tue, 17 Mar 2020 at 22:56, Peter Robinson  wrote:
> > > > I noticed that the kernel builds on the branched fedora32 kernels do
> >
> > not
> > > > have CONFIG_TCP_CONG_BBR set ; meaning no shiny.
> > >
> > > The kernel module is built:
> > > $ modinfo tcp_bbr
> > > filename:
> >
> > /lib/modules/5.6.0-0.rc5.git0.2.fc32.x86_64/kernel/net/ipv4/tcp_bbr.ko.xz
> > > description:    TCP BBR (Bottleneck Bandwidth and RTT)
> > > > root@kuriiti network-scripts]# sysctl
> > > > net.ipv4.tcp_available_congestion_control
> > > > net.ipv4.tcp_available_congestion_control = reno cubic
> > >
> > > If you manually load the module it's there:
> > >
> > > # sysctl -a | grep tcp_available_congestion_control
> > > net.ipv4.tcp_available_congestion_control = reno cubic
> > > # modprobe tcp_bbr
> > > # sysctl -a | grep tcp_available_congestion_control
> > > net.ipv4.tcp_available_congestion_control = reno cubic bbr
> > >
> > > I'm not sure the "official" way to use different congestion control
> > > algorithms but I'm guessing it's something you can use NetworkManager
> > > to specify and autoload them.
> > >
> > > Peter
> >
> > --
> > Joel Wirāmu Pauling
> > Senior Solutions Architect
> > Mobile: [phone]
> > Email: [e-mail] 
> > _______________________________________________
> > kernel mailing list -- [e-mail]
> > To unsubscribe send an email to [e-mail]
> > Fedora Code of Conduct:
> > https://docs.fedoraproject.org/en-US/project/code-of-conduct/
> > List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
> > List Archives:
> > https://lists.fedoraproject.org/archives/list/[e-mail]