Files
punktfunk/.gitea/workflows/windows-drivers.yml
T
enricobuehler 95a08e99c3 feat(host/windows): seal the host↔driver channels (frame + gamepad, proto v2)
Frame ring (pf-vdisplay) and both gamepad SHM channels move off named Global\
objects (openable by any sibling LocalService) to UNNAMED sections/events whose
handles the host DuplicateHandles into the driver's verified WUDFHost with least
access — frame delivery over the SYSTEM+admins-only IOCTL_SET_FRAME_CHANNEL,
pads over a 32-byte named bootstrap mailbox (pid + handle value only, DoS-bounded;
HID minidrivers have no control device). Driver-validated pad_index kills
cross-pad redirects; v1↔v2 mixes fail closed with diagnosis logs on both sides.
Sibling-LocalService denial proven empirically (design/idd-push-security.md,
design/gamepad-channel-sealing.md).

Driver-side raw ops now live behind pf-umdf-util (checked shm accessors, the
forbid(unsafe_code) ChannelClient state machine, WDF request tokens) — the pad
drivers' logic is 100% safe Rust; whole drivers workspace clippy-gated in CI.

driver install --gamepad now sweeps SWD\punktfunk phantom devnodes: a re-created
SwDevice REVIVES the old devnode with its previously-bound driver (never
re-ranks), so an upgrade otherwise leaves the old driver serving — or, across
the v1→v2 fence, a dead pad (found live on the RTX box).

On-glass validated on the RTX 4090 box: frame path 7007 frames p50 2.06 ms
cross-machine; DualSense + XUSB "sealed pad channel mapped"/proto=2 attach via
both the test harness and a real streaming session; phantom-sweep repro.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
2026-07-03 12:08:56 +00:00

163 lines
9.1 KiB
YAML

# Windows driver workspace CI — runs on a self-hosted Windows runner (home-windows-runner-1, host
# mode; label windows-amd64). Part of the Windows-host rewrite (design/windows-host-rewrite.md, M0).
#
# Stage 1 (this file): PROBE the runner's driver toolchain (WDK / EWDK / cargo-make / LLVM / the
# inf2cat/stampinf/devgen/signtool tools) so we know what's provisioned BEFORE writing driver code,
# and build+test the owned ABI crate (pf-driver-proto) on MSVC to prove it compiles cross-OS and the
# CI wiring works. The runner has no RTX GPU — that's fine: builds, the IddCx bindgen/link, the
# /INTEGRITYCHECK self-sign-load, and (later) IDD-push frame flow on the basic display do not need one;
# only live NVENC encode does, which defers to the RTX box.
#
# shell: pwsh deliberately (PowerShell 5.1's Out-File -Encoding utf8 prepends a BOM that corrupts the
# first GITHUB_ENV line — see windows.yml).
name: windows-drivers
on:
workflow_dispatch:
push:
branches: [main]
paths:
- '.gitea/workflows/windows-drivers.yml'
- 'crates/pf-driver-proto/**'
- 'packaging/windows/drivers/**'
pull_request:
paths:
- '.gitea/workflows/windows-drivers.yml'
- 'crates/pf-driver-proto/**'
- 'packaging/windows/drivers/**'
# Driver builds need the WDK on the runner - the driver-build job below self-provisions it via
# scripts/ci/ensure-windows-toolchain.ps1, a fast no-op once already present.
jobs:
probe-and-proto:
runs-on: windows-amd64
timeout-minutes: 30
defaults:
run:
shell: pwsh
steps:
- uses: actions/checkout@v4
- name: Probe driver toolchain (informational — never fails the job)
continue-on-error: true
run: |
$ErrorActionPreference = 'Continue'
function head($t) { Write-Host ""; Write-Host "===== $t =====" }
head "Windows Kits roots"
$kits = @('C:\Program Files (x86)\Windows Kits\10', 'C:\Program Files\Windows Kits\10')
foreach ($k in $kits) { if (Test-Path $k) { Write-Host "found: $k" } }
head "SDK Include versions (um vs km — km => WDK present)"
foreach ($k in $kits) {
$inc = Join-Path $k 'Include'
if (Test-Path $inc) {
Get-ChildItem $inc -Directory | ForEach-Object {
$hasUm = Test-Path (Join-Path $_.FullName 'um')
$hasKm = Test-Path (Join-Path $_.FullName 'km')
$wdf = Test-Path (Join-Path $_.FullName 'km\wdf\umdf\2.31')
$iddcx = (Get-ChildItem (Join-Path $_.FullName 'um\iddcx') -Directory -ErrorAction SilentlyContinue | ForEach-Object { $_.Name }) -join ','
Write-Host ("{0,-16} um={1,-5} km={2,-5} wdf2.31={3,-5} iddcx=[{4}]" -f $_.Name, $hasUm, $hasKm, $wdf, $iddcx)
}
}
}
head "Driver tooling (inf2cat / stampinf / signtool / devgen / InfVerif)"
foreach ($tool in 'inf2cat.exe','stampinf.exe','signtool.exe','devgen.exe','InfVerif.exe','makecat.exe') {
$hits = @()
foreach ($k in $kits) {
$hits += Get-ChildItem -Path $k -Filter $tool -Recurse -ErrorAction SilentlyContinue |
Where-Object { $_.FullName -match '\\x64\\' } | Select-Object -First 1 -ExpandProperty FullName
}
$hits = $hits | Where-Object { $_ } | Select-Object -First 1
Write-Host ("{0,-14} -> {1}" -f $tool, ($(if ($hits) { $hits } else { 'NOT FOUND' })))
}
head "EWDK"
Write-Host ("EWDKROOT = " + ($env:EWDKROOT ?? '<unset>'))
head "LLVM / clang (bindgen 0.72 builds on the runner default clang)"
Write-Host ("LIBCLANG_PATH = " + ($env:LIBCLANG_PATH ?? '<unset>'))
$clang = Get-Command clang -ErrorAction SilentlyContinue
if ($clang) { & clang --version } else { Write-Host "clang: NOT on PATH" }
head "cargo-make (the gamepad drivers' build driver)"
$cm = & cargo make --version 2>&1; Write-Host $cm
head "Rust + targets"
& rustc -V; & cargo -V
Write-Host "installed targets:"; & rustup target list --installed
head "Env knobs the WDK build cares about"
Write-Host ("Version_Number = " + ($env:Version_Number ?? '<unset>'))
Write-Host ("CARGO_HOME = " + ($env:CARGO_HOME ?? '<unset>'))
Write-Host ("CARGO_TARGET_DIR (daemon) = " + ($env:CARGO_TARGET_DIR ?? '<unset>'))
- name: Build + test pf-driver-proto (MSVC)
run: |
# Short target dir to dodge MAX_PATH inside the deep act host workdir (see windows.yml).
$env:CARGO_TARGET_DIR = "C:\t\drv"
cargo build -p pf-driver-proto
cargo test -p pf-driver-proto
cargo clippy -p pf-driver-proto --all-targets -- -D warnings
cargo fmt -p pf-driver-proto -- --check
# Build the UMDF driver workspace (wdk-probe) on windows-drivers-rs: proves wdk-sys bindgen/link works
# on the runner's WDK + LLVM, that pf-driver-proto path-deps into a driver, and exposes the produced
# DLL's FORCE_INTEGRITY (/INTEGRITYCHECK) bit — the M0 self-signed-load question.
driver-build:
runs-on: windows-amd64
timeout-minutes: 45
defaults:
run:
shell: pwsh
# In-tree target dir on purpose: wdk-build's find_top_level_cargo_manifest() walks UP from OUT_DIR
# to the first ancestor with a Cargo.lock, so a relocated CARGO_TARGET_DIR (C:\t\…) hides the
# workspace lock and it panics. The driver deps have no deep CMake-from-source crates, so the
# default in-tree target stays well under MAX_PATH (unlike the SDL3/audiopus client build).
working-directory: packaging/windows/drivers
env:
# wdk-build otherwise picks 10.0.28000.0 (no km/crt) and bindgen fails — pin the WDK SDK version.
Version_Number: '10.0.26100.0'
# No LIBCLANG_PATH pin: the vendored bindgen 0.72 builds clean on the runner's default clang 22
# (the shipping pack proves it). A 0.71-era layout-test overflow once needed LLVM 21; the 0.72 bump
# retired that — see design/windows-build-and-packaging.md.
steps:
- uses: actions/checkout@v4
- name: Ensure Windows toolchain (WDK, FFmpeg, Inno Setup, ARM64 target)
# Shared self-provision step (also used by windows.yml/windows-msix.yml/windows-host.yml) so
# driver-build is self-sufficient on any windows-amd64 runner and never races a manually
# dispatched provisioning workflow landing on a different one. Path is relative to the job
# working-directory (packaging/windows/drivers). Near-noop once the toolchain is present.
run: ../../../scripts/ci/ensure-windows-toolchain.ps1
- name: cargo build the driver workspace (wdk-probe + wdk-iddcx + pf-vdisplay + gamepad drivers)
# Whole workspace: wdk-probe (toolchain/surface-assert probe) + wdk-iddcx (DDI wrappers) +
# pf-vdisplay (the real IddCx driver) + pf-umdf-util (the safe UMDF primitive layer) + the two
# gamepad drivers. pf-vdisplay linking proves the IddCx call sites resolve against IddCxStub
# end-to-end (M1 step 2 gate); the gamepad drivers prove pf-umdf-util's WDF dispatch links.
run: cargo build -v
- name: cargo clippy the shipped drivers (-D warnings — enforces the unsafe-audit gates)
# The gamepad drivers' business logic is 100% safe (it moved onto pf-umdf-util, the audited
# unsafe layer); pf-vdisplay + wdk-iddcx are inherently FFI-bound but every `unsafe {}` carries a
# `// SAFETY:` proof. Both invariants are lint-gated (`unsafe_op_in_unsafe_fn` +
# `undocumented_unsafe_blocks`); this step keeps them from regressing. (wdk-probe is a
# toolchain-only probe crate and is excluded.)
run: cargo clippy -p pf-umdf-util -p pf-xusb -p pf-dualsense -p wdk-iddcx -p pf-vdisplay --all-targets -- -D warnings
- name: cargo fmt --check the safe-layer + gamepad drivers
run: cargo fmt -p pf-umdf-util -p pf-xusb -p pf-dualsense --check
- name: Inspect /INTEGRITYCHECK (before) — expect FORCE_INTEGRITY set by wdk-build
run: |
# explicit --target (.cargo/config.toml) -> output under the triple subdir.
$dll = "target\x86_64-pc-windows-msvc\debug\pf_vdisplay.dll"
if (-not (Test-Path $dll)) { throw "pf_vdisplay.dll not produced at $dll" }
$b = [IO.File]::ReadAllBytes($dll)
$pe = [BitConverter]::ToInt32($b, 0x3c)
$dllchar = [BitConverter]::ToUInt16($b, $pe + 0x5e) # OptionalHeader.DllCharacteristics
Write-Host ("pf_vdisplay.dll built OK ({0:N0} bytes)" -f (Get-Item $dll).Length)
Write-Host ("BEFORE: DllCharacteristics = 0x{0:X4}; FORCE_INTEGRITY = {1}" -f $dllchar, (($dllchar -band 0x0080) -ne 0))
- name: Clear FORCE_INTEGRITY (self-signed-load fix) + verify
# wdk-build sets /INTEGRITYCHECK unconditionally -> a self-signed driver won't load. Clear the PE
# bit deterministically (the reusable packaging step; signing/.cat happen later for real drivers).
run: ../clear-force-integrity.ps1 -Path target\x86_64-pc-windows-msvc\debug\pf_vdisplay.dll