fix(client/pkg): ship 32MB UDP recv-buffer sysctl with the Linux client
ci / web (push) Successful in 27s
ci / docs-site (push) Successful in 31s
apple / swift (push) Successful in 1m16s
ci / rust (push) Successful in 2m6s
ci / bench (push) Successful in 1m36s
docker / build-push (--build-arg FEDORA_VERSION=44, ci, ci/fedora-rpm.Dockerfile, punktfunk-fedora44-rpm) (push) Successful in 6s
docker / build-push (., web/Dockerfile, punktfunk-web) (push) Successful in 5s
docker / build-push (ci, ci/fedora-rpm.Dockerfile, punktfunk-fedora-rpm) (push) Successful in 3s
docker / build-push (ci, ci/rust-ci.Dockerfile, punktfunk-rust-ci) (push) Successful in 4s
docker / build-push (docs-site, docs-site/Dockerfile, punktfunk-docs) (push) Successful in 4s
deb / build-publish (push) Successful in 2m19s
rpm / build-publish (bazzite, punktfunk-fedora-rpm) (push) Successful in 4m51s
docker / deploy-docs (push) Successful in 17s
rpm / build-publish (fedora-44, punktfunk-fedora44-rpm) (push) Successful in 4m24s

The client asks the kernel for a 32 MB SO_RCVBUF, but the kernel silently clamps
it to net.core.rmem_max — whose default is far too small. A too-small recv buffer
is the dominant client-side wall above ~1 Gbps. Measured live (Fedora host -> two
clients, real 2.5G LAN, GSO off): a client capped at 4 MB rmem_max dropped 31.6%
of a 2 Gbps stream at the receiver, while a 32 MB client delivered the same
2 Gbps at 0.0% loss. The host already shipped this tuning; the client packages
didn't (the RPM's %post even referenced the host-only file), so a client-only
install streamed lossy at high bitrate.

Add scripts/99-punktfunk-client-net.conf (rmem/wmem = 32 MB, distinct filename so
host+client can coexist) and ship+apply it from both the .deb (build-client-deb.sh)
and the RPM client subpackage (install, %files client, %post client).

For reference the full ladder (punktfunk speed-test): 0% loss to 1.5 Gbps on a
4 MB client; 31.6% at 2 Gbps on 4 MB vs 0% at 2 Gbps on 32 MB. iperf3 put the raw
link at ~2.35 Gbps TCP / ~2.4 Gbps UDP, so the stack now tracks the wire given a
big enough recv buffer.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-06-14 08:45:19 +00:00
parent 4d26f61e40
commit 9f92dc505b
3 changed files with 25 additions and 0 deletions
+10
View File
@@ -154,6 +154,12 @@ install -Dm0644 packaging/linux/io.unom.Punktfunk.desktop \
# DualSense hidraw access (full pad fidelity through SDL's HIDAPI driver).
install -Dm0644 scripts/70-punktfunk-client.rules \
%{buildroot}%{_udevrulesdir}/70-punktfunk-client.rules
# UDP receive-buffer tuning (32 MB) — the client asks for a 32 MB SO_RCVBUF; without raising
# net.core.rmem_max the kernel clamps it and high-bitrate streams overflow at the receiver
# (measured: 4 MB cap = 31.6% loss at 2 Gbps, 32 MB = 0%). Distinct filename from the host's so
# both can be installed on one box.
install -Dm0644 scripts/99-punktfunk-client-net.conf \
%{buildroot}%{_prefix}/lib/sysctl.d/99-punktfunk-client-net.conf
# Headless session helpers + example config + OpenAPI doc (reference material).
install -d %{buildroot}%{_datadir}/%{name}/headless
@@ -184,12 +190,16 @@ install -Dm0644 docs/api/openapi.json %{buildroot}%{_datadir}/%
%{_bindir}/punktfunk-client
%{_datadir}/applications/io.unom.Punktfunk.desktop
%{_udevrulesdir}/70-punktfunk-client.rules
%{_prefix}/lib/sysctl.d/99-punktfunk-client-net.conf
%post client
# Pick up the DualSense hidraw rule without a reboot (best-effort; on rpm-ostree it
# applies on the next boot into the layered deployment).
udevadm control --reload-rules 2>/dev/null || :
udevadm trigger --subsystem-match=hidraw 2>/dev/null || :
# Apply the UDP recv-buffer tuning now (also auto-applied at boot by systemd-sysctl; on
# rpm-ostree it takes effect on the next boot into the layered deployment).
sysctl -p %{_prefix}/lib/sysctl.d/99-punktfunk-client-net.conf >/dev/null 2>&1 || :
%post
# Reload udev so /dev/uinput picks up the new rule without a reboot (best-effort).