floatdrop / zoxy / 28693749788
96%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 25
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 96.824% (+0.07%) from 96.752%
28693749788

push

github

floatdrop 
io: kTLS seam — enable_kernel_tls + a sendmsg ring op (kTLS slice 1)

src/tls/kernel.zig mirrors the linux/tls.h uapi surface by hand,
verified against linux-headers 6.18.7 with comptime size/offset
asserts: crypto-info structs for AES-GCM-128/256, record types, and
RecordTypeControl — one TLS_SET_RECORD_TYPE cmsg laid out for
msghdr.control (CMSG_SPACE checked at comptime).

IO.enable_kernel_tls (sync seam, like shutdown_socket) attaches the
"tls" ULP and installs one crypto-info per direction; failure means
"stay on the userspace relay". The simulation backend always refuses
(the virtual kernel has no ULP), so a simulated proxy can never
switch. io.send_message is a sendmsg(2) ring op — caller owns the
msghdr and everything it references until completion — which the sim
maps onto a plain send of the single segment.

Tests (skip cleanly when the tls module cannot load — CI runners):
records round-trip between a TLS_TX and TLS_RX loopback socket pair
with distinct per-direction keys (real kernel crypto, no OpenSSL);
a close_notify alert sent through the ring arrives typed via
recvmsg cmsg (TLS_GET_RECORD_TYPE = alert); and a control record
hitting a plain read fails with EIO — the exact contract the §6
design relies on for post-switch KeyUpdate handling.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

5244 of 5416 relevant lines covered (96.82%)

0.97 hits per line

Coverage Regressions

Lines Coverage ∆ File
25
94.13
 0.0% io/test_io.zig
10
95.59
 -0.61% io/linux.zig
Jobs
ID Job ID Ran Files Coverage
1 28693749788.1 25
96.82
 GitHub Action Run
Source Files on build 28693749788