chromium-fourier r2 + firefox-fourier 150.0.1 + KWIN_PIVOT.md

chromium-fourier:
- patch 3/3 nv12-external-oes-on-modifier-external-only.patch — adds
  NativePixmapEGLBinding::ModifierRequiresExternalOES helper, extends
  OzoneImageGLTexturesHolder::GetBinding to honor EGL external_only
  flag for NV12 dmabufs on panfrost / panthor. Validated on ohm
  (RK3566 hantro mainline 6.19.10): bbb_1080p30_h264.mp4 plays at
  34.7 % combined CPU vs ~131 % pre-patch baseline (~3.8x).
- PKGBUILD pkgrel 1->2, source array + sha256sums + prepare() hook for
  patch 4, patch numbering 1/2,2/2 -> 1/3,2/3,3/3.
- NEXT.md appended with 2026-04-28 section: patch 4 design, validation
  log, KWin GL_ALPHA bug pinpoint (preexisting since 2026-03-06,
  affects every wayland video client; unrelated to chromium-fourier),
  device-renumbering note (/dev/video1 = encoder post-reboot).
- KWIN_PIVOT.md: 4-phase plan to identify and patch KWin's
  glTexImage2D(internalFormat=GL_ALPHA) site, ohm-only test plan,
  scope discipline.
- patches/ now tracked (compiler-rt-adjust-paths, enable-v4l2,
  wayland-allow-direct-egl-gles2, nv12-external-oes); the dead-end
  chromeos-pipeline-bypass.patch removed.

firefox-fourier:
- 4 patches (gfxinfo v4l2 stateless fourccs, libwrapper hwdevice ctx,
  ffmpegvideo v4l2-request route, prefs v4l2-request default).
- PKGBUILD bumped to firefox 150.0.1, Arch toolchain glue patches
  layered in, mozconfig with --without-wasm-sandboxed-libraries for
  ALARM, package() launcher fix (rm -f symlink before cat > to avoid
  ENOENT through the dangling /usr/local symlink mach install drops).
- 150.0.1-1-aarch64.pkg.tar.zst built on boltzmann (95 MB), pending
  fresnel power-on for V4L2 stateless validation on RK3399.

arch/chromium-fourier/KWIN_PIVOT.md

@@ -0,0 +1,181 @@
+# KWin pivot — fix the `glTexImage2D(GL_ALPHA)` stall
+
+## What we know
+
+KWin 6.6.4-1 on Arch Linux ARM (Plasma 6.6.4-1, mesa 26.0.5-1, libdrm
+2.4.131-1) on ohm (PineTab2 / RK3566 / panfrost) silently corrupts its
+GL command queue mid-frame whenever a wayland client posts a video
+buffer. The journal carries a rolling stream of:
+
+```
+kwin_wayland: 0x4: GL_INVALID_VALUE in glTexImage2D(internalFormat=GL_ALPHA)
+kwin_wayland: 0x4: GL_INVALID_OPERATION in glTexSubImage2D(invalid texture level 0) × N
+```
+
+`GL_ALPHA` is not a valid `internalFormat` for `glTexImage2D` under
+**OpenGL ES 3.x** (it was the GLES1.x single-channel alpha format;
+GLES3 deprecates it for sized formats — `GL_R8`, `GL_LUMINANCE8_ALPHA8`,
+etc.). Once the texture allocation fails, the `glTexSubImage2D` calls
+that should populate it all error at level 0. KWin keeps retrying the
+same broken upload every frame, never recovers, and the present-callback
+path that depends on that texture stops acking client frames. Every
+wayland video client deadlocks on the missing ack.
+
+First occurrence in this box's journal: **2026-03-06** — the bug
+predates any chromium-fourier work by roughly seven weeks.
+
+## Triangulation already in hand
+
+| Client | Outcome |
+|---|---|
+| chromium-fourier 149-r2 (with patch 3/3) | plays ~3 s @ 34.7 % CPU then renderer/GPU park in `futex_do_wait` |
+| chromium-fourier 149-r2 (without patch 3/3) | plays ~10 s (slower path delays surfacing) then identical deadlock |
+| VLC | `cannot convert decoder/filter output to any format supported by the output` → `could not initialize video chain` |
+| mpv `--vo=null --hwdec=v4l2request` | `Could not create device.` (mpv-side bug, separate, unrelated) |
+| ffmpeg `-hwaccel v4l2request -i bbb -f null -` | plays through clean at 36 fps; hardware path is healthy |
+
+Decode path is healthy on this hardware. The wall is exclusively the
+compositor's GL backend.
+
+## Constraint: ohm is the only test box on hand
+
+ampere (RK3588 / panthor) is in the boxes-from-Shenzhen pile, currently
+DOWN. fresnel (RK3399 / Pinebook Pro) is offline. boltzmann (Rock 5
+ITX+ build host) doesn't run KWin. We do every step on ohm; we accept
+the wifi flakiness and the occasional reboot.
+
+## Phase 1 — Reproduce outside chrome and bound the trigger (1 evening)
+
+Goal: a deterministic, headless-or-near-headless reproduction that
+doesn't require launching a 800-MB browser.
+
+1. **Smallest-possible client.** Build a 50-line C wayland client that
+   creates a `wp_linux_dmabuf_v1` buffer, pumps frames at 30 fps, and
+   exits when KWin first errors. Use `weston-simple-dmabuf-egl` from
+   the `weston` package as a starting template — already does exactly
+   this but without our specific format/modifier matrix.
+2. **Vary the format/modifier matrix.** Run the smallest-possible
+   client with each of: NV12 + LINEAR, NV12 + AFBC, NV12 + AFRC,
+   AR24 + LINEAR, XR24 + LINEAR. We already know NV12 paths trigger;
+   confirming AR24/XR24 do *not* trigger localizes the bug to KWin's
+   YUV import path (vs a generic dmabuf import bug).
+3. **Vary the buffer dimensions.** Some KWin texture-cache paths
+   allocate fixed-size internal scratch textures; non-power-of-two,
+   non-multiple-of-16, or specifically odd-aspect cases sometimes
+   trigger paths that healthy aspect ratios skip. Test 1920×1080,
+   1280×720, 854×480, 640×360 and a deliberately weird 1366×768.
+4. **Vary KWin scene type.** Switch
+   `kwin_wayland --scene-type=opengl` vs `--scene-type=opengl-es`
+   (current default on this hardware). If the bug only fires under
+   GLES, that's a strong signal — the offending site is in a
+   GLES-only fallback.
+
+By the end of Phase 1 we should have a one-line `weston-simple-dmabuf-egl
+-format=NV12 -modifier=…` that triggers the GL_ALPHA error within
+seconds, plus a yes/no answer to "does AR24 also trigger".
+
+## Phase 2 — Identify the call site (1–2 evenings)
+
+The crime scene is somewhere in `kwin/src/scene/*` or
+`kwin/src/effects/*`. Suspects, ranked:
+
+- **`SurfaceItemWayland::createPixmapTexture` → `GLTexture::create`
+  with `GL_ALPHA`.** This is the most likely path: KWin allocates a
+  fallback per-plane texture when the dmabuf import path can't take
+  the buffer whole. NV12 has a Y plane (single-channel) and a CbCr
+  plane (two-channel); historically the Y plane has been allocated as
+  `GL_ALPHA` in software fallbacks. If the EGL dmabuf import returned
+  `EGL_BAD_ATTRIBUTE` for `external_only` modifiers and KWin fell
+  through to per-plane, this is exactly where it would land.
+- **`BlurEffect::initBlurTexture` / `BackgroundContrastEffect::*`.**
+  Single-channel noise textures for blur dither. Less likely (these
+  fire on every frame regardless of video clients) but listed for
+  completeness.
+- **Window-decoration text glyph cache.** Qt's QGLTexture historically
+  requested `GL_ALPHA` for monochrome glyph atlases. Plasma 6 should
+  have moved to `GL_RED` long ago, but a stale code path in a
+  third-party theme or systray icon could still hit it.
+- **Cursor texture upload via `wl_shm_pool` + ARGB8888.** KWin's
+  cursor scene sometimes uploads via glTexImage2D — but the format
+  there is `GL_RGBA`, not `GL_ALPHA`. Probably not the suspect.
+
+Tooling to identify *which*:
+
+1. **`apitrace trace --api egl kwin_wayland …`** then
+   `apitrace dump trace.trace | grep -B5 GL_ALPHA`. Apitrace gives
+   us the C++ call stack at the offending site if KWin was built with
+   debug symbols.
+2. **`MESA_GL_DEBUG=context KWIN_GL_DEBUG=1 kwin_wayland --replace`**
+   plus `glDebugMessageCallback` already installed in KWin's
+   `OpenGLBackend` will print the source/type/severity for each
+   `GL_INVALID_VALUE`. Whether the file/line in the message includes
+   the user-space caller depends on Mesa's debug-extension support;
+   on panfrost it usually does include the GL function name and an
+   ID, but not the C++ source — that is what apitrace adds.
+3. **Build kwin from source** (`extra/kwin` PKGBUILD on Arch ARM,
+   patch in `-DDEBUG=ON`, `-DCMAKE_BUILD_TYPE=Debug`) so the call
+   stacks resolve to file:line.
+
+## Phase 3 — Write the patch (½ evening once Phase 2 is done)
+
+If the offender is a `GL_ALPHA` allocation in a GLES3 context, the
+fix is mechanical:
+
+```diff
+- glTexImage2D(GL_TEXTURE_2D, 0, GL_ALPHA, width, height, 0,
+-              GL_ALPHA, GL_UNSIGNED_BYTE, data);
++ glTexImage2D(GL_TEXTURE_2D, 0, GL_R8, width, height, 0,
++              GL_RED, GL_UNSIGNED_BYTE, data);
+```
+
+…and adjust the consuming shader's swizzle:
+
+```diff
+- gl_FragColor = vec4(texture2D(s, uv).a, …);
++ gl_FragColor = vec4(texture2D(s, uv).r, …);
+```
+
+If the offender is a per-plane fallback in the dmabuf import path
+(suspect #1 above), the patch is larger because the right fix is to
+*not fall through to the broken path* — handle the `external_only`
+case by binding `GL_TEXTURE_EXTERNAL_OES` instead. That mirrors the
+chromium-fourier patch 3/3 done at the chromium layer; symmetry says
+KWin should do the same in its `glTexImage` consumer.
+
+## Phase 4 — Ship and upstream (1 evening)
+
+1. **Local Arch package** as `kwin-fourier` under
+   `marfrit-packages/arch/kwin-fourier/`, sibling to chromium-fourier
+   and firefox-fourier. PKGBUILD inherits from `extra/kwin`, drops
+   in our patch, bumps `pkgrel`. Same `provides=kwin conflicts=kwin`
+   pattern.
+2. **Validate on ohm** by running the chromium-fourier 149-r2 build +
+   the bbb sample for a minute uninterrupted. Success = no GL_ALPHA
+   in the journal, no stall, smooth playback at the 34.7 % CPU
+   number from the chromium validation.
+3. **Upstream** via:
+   - File a `kwin` bug on bugs.kde.org with: apitrace fragment, our
+     hardware (Mali-G52 panfrost on RK3566 mainline), exact mesa
+     version, repro steps via `weston-simple-dmabuf-egl` if Phase 1
+     produced one.
+   - Push an MR to invent.kde.org/plasma/kwin against `master`.
+4. **Document** the fix in `chromium-fourier/docs/dmabuf-zero-copy.md`
+   so the next person who lands on the same wall finds the breadcrumb
+   trail.
+
+## What success looks like
+
+`chromium-fourier-149-r2` on ohm under KWin Wayland plays
+`bbb_1080p30_h264.mp4` end-to-end at the 34.7 % CPU figure already
+recorded by the architectural validation, with zero `GL_INVALID_VALUE`
+in the journal during playback. That number is the goal of the entire
+chromium-fourier campaign for RK3566 — it is currently blocked on a
+bug that has nothing to do with chromium.
+
+## Scope discipline
+
+We do not turn this into "audit the entire KWin GLES backend." If
+Phase 2 surfaces additional latent GL_INVALID_* errors that don't
+matter for video playback, we note them in the bug report and move
+on. The pivot is explicitly "remove this single wall so the
+chromium-fourier patch series can ship a working stack on RK3566."

arch/chromium-fourier/NEXT.md

@@ -148,3 +148,281 @@ to install chromium's bundled clang (x86_64 host, arm64 sysroot), then
 The boltzmann `chromium-builder` LXD container is preserved as fallback
 but no longer the active build host. If cross-compile pans out, that
 container can be torn down.
+
+## First runtime validation on ohm — 2026-04-26 22:26 UTC
+
+Cross-compile produced a working aarch64 binary (chrome 647 MB ELF +
+chrome_crashpad_handler 4.3 MB + .pak + locales). Tarball
+`chromium-fourier-147.tar.gz` (226 MB) transferred CT 220 → hertz → ohm.
+Launched in mfritsche's KWin Wayland session (tty2, panfrost render
+node) playing `bbb_1080p30_h264.mp4` from file:// with
+`LIBVA_DRIVER_NAME=v4l2_request`,
+`LIBVA_V4L2_REQUEST_VIDEO_PATH=/dev/video0`,
+`--use-gl=egl --ozone-platform=wayland
+--enable-features=VaapiVideoDecodeLinuxGL,AcceleratedVideoDecodeLinuxGL
+--disable-features=UseChromeOSDirectVideoDecoder
+--autoplay-policy=no-user-gesture-required`.
+
+**Result: V4L2 path NOT engaged.** Chrome 147 routes the H.264 stream
+through `MojoVideoDecoderService` → `media/filters/ffmpeg_video_decoder.cc`
+(software FFmpeg). Renderer pegs at ~92 % CPU, `/dev/video0` is never
+opened (`fuser` returns empty), no `V4L2VideoDecoder` /
+`VaapiVideoDecoder` log lines appear at `--v=1
+--vmodule="*/vaapi/*=2,*/v4l2/*=2,*video_decoder*=2,*media/gpu/*=2"`.
+Compositor also fell back to software (`Switching to software
+compositing.` even though panfrost render node was picked) — secondary
+issue, separate from the codec wall.
+
+**Conclusion**: 7Ji-style gn args (`use_v4l2_codec=true
+use_v4lplugin=true use_linux_v4l2_only=true`) alone are insufficient
+on chromium 147. The V4L2VideoDecoder factory is still gated behind
+`BUILDFLAG(IS_CHROMEOS)` — `media/mojo/services/gpu_mojo_media_client_*.cc`
+and `media/gpu/gpu_video_decode_accelerator_factory.cc` only register
+the V4L2 path on ChromeOS targets.
+
+## Validation pass 2 — 2026-04-26 22:38 UTC — V4L2VDA proven engaged
+
+Two distinct issues were diagnosed and the codec one was fully resolved
+without source surgery beyond a 2-line patch:
+
+### Issue 1 — runtime master gate
+`media::kAcceleratedVideoDecodeLinux` (user-visible feature name
+"AcceleratedVideoDecoder") is hard-coded in
+`media/base/media_switches.cc:750` to `FEATURE_ENABLED_BY_DEFAULT` only
+when `BUILDFLAG(USE_VAAPI)` is set. On a USE_V4L2_CODEC-only build it
+defaults DISABLED, the linux gpu_mojo_media_client returns
+`VideoDecoderType::kUnknown`, and chrome silently falls back to
+`media/filters/ffmpeg_video_decoder.cc`.
+
+**Fix**: 2-line patch (now `patches/enable-v4l2-decoder-default.patch`):
+```
+-#if BUILDFLAG(USE_VAAPI)
++#if BUILDFLAG(USE_VAAPI) || BUILDFLAG(USE_V4L2_CODEC)
+```
+
+The placeholder `chromeos-pipeline-bypass.patch` was deleted; PKGBUILD
+now references the real patch. **Verified to apply cleanly on the CT 220
+tree** (chromium 149 main).
+
+### Issue 2 — bundled GL libs missing from tarball
+The first runtime tarball shipped only `chrome` + `.pak` + locales +
+`chrome_crashpad_handler`. It omitted `libEGL.so` / `libGLESv2.so`
+(ANGLE) plus `libvk_swiftshader.so` and `libvulkan.so.1`. Without these,
+the GPU process logs `gl::init::InitializeStaticGLBindingsOneOff failed`
+and chrome falls into "Switching to software compositing." mode — which
+*also* gates the V4L2 path off because the gpu_mojo_media_client never
+gets a chance to dispatch.
+
+Additionally, `--use-gl=egl` is rejected ("Requested GL implementation
+gl=egl-gles2,angle=none not found in allowed implementations:
+[(gl=egl-angle,angle=opengl|opengles|vulkan)]"): the build only allows
+ANGLE-mediated paths. Right launcher invocation:
+`--use-gl=angle --use-angle=gles`.
+
+**Fix**: package the four libs alongside `chrome` and update the
+launcher flag set. Both will be encoded in the next iteration of the
+PKGBUILD's `package()` and a `chromium-fourier` launcher script.
+
+### What we observed once both fixes were in place
+With patch + bundled libs + `--enable-features=AcceleratedVideoDecoder`
++ `--use-gl=angle --use-angle=gles`, chrome on RK3566 hantro logs:
+
+```
+[gpu]: V4L2VideoDecoder()
+[gpu]: Open(): No devices supporting H264 for type: 0     <- type=0 is single-planar; chrome retries multi-planar
+[gpu]: InitializeBackend(): Using a stateless API for profile: h264 main and fourcc: S264
+[gpu]: SetupInputFormat(): Input (OUTPUT queue) Fourcc: S264
+[gpu]: AllocateInputBuffers(): Requesting: 17 OUTPUT buffers of type V4L2_MEMORY_MMAP
+[gpu]: SetExtCtrlsInit(): Setting EXT_CTRLS for H264
+[gpu]: SetupOutputFormat(): Output (CAPTURE queue) candidate: NV12
+[gpu]: ContinueChangeResolution(): Requesting: 6 CAPTURE buffers of type V4L2_MEMORY_MMAP
+[renderer]: OnDecoderSelected<video>: V4L2VideoDecoder
+MediaEvent: "Selected V4L2VideoDecoder for video decoding,
+             config: codec: h264, profile: h264 main, [...]
+             coded size: [1920,1080], visible rect: [0,0,1920,1080]"
+```
+
+`fuser /dev/video1 /dev/media0` shows `chrome` (gpu pid) holding both
+fds. The hantro stateless decoder is engaged. **First end-to-end
+chromium-fourier V4L2 hardware decode validation: PASS** for H.264
+1080p Big Buck Bunny on PineTab2.
+
+Caveat: the render-side CPU was still ~85% during playback. Subsequent
+investigation traced this to a different root cause than initially
+guessed (see Pass 3 below).
+
+## Validation pass 3 — 2026-04-26 22:50 UTC — zero-copy diagnosis
+
+The 85 % CPU is **not** caused by software compositing or dmabuf v5
+negotiation. The dmabuf-v5 warning ("Binding to zwp_linux_dmabuf_v1
+version 4 but version 5 is available") is benign — chrome happily binds
+to its supported max (v4). The `WaylandZwpLinuxDmabuf::OnTrancheFlags`
+NOTIMPLEMENTED is also benign — KWin sends it, chrome ignores it, but
+the substantive feedback (formats + modifiers) lands via
+`OnTrancheFormats` / `OnTrancheTargetDevice` regardless.
+
+Real cause: `gpu_feature_info.supports_nv12_gl_native_pixmap` ends up
+**false** on this build. With it false, V4L2-decoded NV12 frames go
+through the NV12-to-AR24 VPP conversion path (see
+`media/mojo/services/gpu_mojo_media_client_linux.cc`
+`GetPreferredRenderableFourccs` — without NV12 native pixmap support,
+only `Fourcc::AR24` is added to the renderable set, forcing the VPP).
+That's where the 85 % is spent.
+
+Why is `supports_nv12_gl_native_pixmap` false?
+`GLOzoneEGLWayland::CanImportNativePixmap` (in
+`ui/ozone/platform/wayland/gpu/wayland_surface_factory.cc`) requires the
+chrome GL display to expose `EGL_EXT_image_dma_buf_import`. With
+`--use-gl=angle --use-angle=gles`, chrome's GL display sits behind
+ANGLE's EGL, and ANGLE's GLES backend on Linux does not propagate
+`EGL_EXT_image_dma_buf_import` from the underlying mesa EGL up to its
+clients. Verified directly: `EGL_PLATFORM=surfaceless eglinfo` on ohm
+shows panfrost native EGL exposes both
+`EGL_EXT_image_dma_buf_import` and `EGL_EXT_image_dma_buf_import_modifiers` —
+the capability is there at the panfrost layer, ANGLE just hides it.
+
+We tried `--use-gl=egl` (direct EGL/GLES2, bypass ANGLE) but were
+rejected with "Requested GL implementation (gl=egl-gles2,angle=none) not
+found in allowed implementations". `WaylandSurfaceFactory::GetAllowedGLImplementations()`
+in chromium 149 advertises only ANGLE-mediated impls; the
+`kGLImplementationEGLGLES2` slot is missing from the list. The
+`CreateViewGLSurface` dispatcher does still handle that impl — only the
+*advertisement* was tightened.
+
+### Patch 2/2 — `wayland-allow-direct-egl-gles2.patch`
+3-line diff in `ui/ozone/platform/wayland/gpu/wayland_surface_factory.cc`:
+```
++    impls.emplace_back(gl::GLImplementationParts(gl::kGLImplementationEGLGLES2));
+     impls.emplace_back(gl::ANGLEImplementation::kOpenGL);
+```
+Re-allows the direct EGL/GLES2 path, ahead of the ANGLE entries so
+chrome picks it by default. Verified to apply cleanly on the CT 220
+tree; staged via `patch -p1` mid-rebuild (ninja's mtime-based rebuild
+will pick up the change automatically).
+
+### Outstanding for next pass (revised)
+1. Rebuild lands → repackage with all four GL libs +
+   `chrome_crashpad_handler` + chrome → ship to ohm.
+2. Validate via `chrome --use-gl=egl --ozone-platform=wayland`
+   `--enable-features=AcceleratedVideoDecoder` (no ANGLE shim) and
+   confirm `chrome://gpu` reports `Native GpuMemoryBuffers: true` and
+   `supports_nv12_gl_native_pixmap=true`. Target CPU during 1080p30 H.264
+   playback: under 30 % combined renderer + gpu.
+3. If (2) passes, declare V1 of chromium-fourier shippable on ohm.
+4. Add a `chromium-fourier` launcher shim under `/usr/bin/` that
+   defaults to `--use-gl=egl --ozone-platform=wayland`.
+5. Sort the chromium 147 vs 149 confusion — the fetch went to ToT on
+   main rather than the 147 release branch. Either pin the branch or
+   accept that we're tracking ToT (probably preferable for V4L2 fixes
+   that are still in flight upstream).
+6. Replicate end-to-end on RK3588 (ampere CoolPi or boltzmann Rock 5
+   ITX+) once the mainline VDPU381 driver is stable on those — those
+   boxes use **panthor** for Mali-G610 (Valhall), not panfrost; the
+   patches should be backend-agnostic but the validation is per-box.
+
+### State of the build host (post pass 3)
+- CT 220 `/build/chromium/src` patched with both
+  `enable-v4l2-decoder-default.patch` and
+  `wayland-allow-direct-egl-gles2.patch` (applied directly with
+  `patch -p1` mid-rebuild; ninja picks up the mtime change).
+- `chromium-rebuild.service` running as a transient unit, output in
+  `/tmp/chromium-rebuild.log`. Most of the 93k ninja steps are cache
+  hits; only the patched files + their downstream objects need
+  recompiling.
+- Tarball still on CT 220 at `/build/chromium-fourier-147.tar.gz`
+  (misleadingly named: it's actually 149.0.7812.0 from the main fetch,
+  not the 147 release tarball — separate cleanup for next pass) and on
+  hertz at `/tmp/chromium-fourier-147.tar.gz`. **Will be replaced by
+  the post-rebuild tarball once it lands.**
+
+---
+
+## 2026-04-28 — Patch 4 lands, KWin owns the residual stall
+
+### Patch 4 — `nv12-external-oes-on-modifier-external-only.patch`
+On panfrost / panthor, every NV12 modifier (LINEAR + AFBC ×2 + AFRC) is
+flagged `external_only` in `eglQueryDmaBufModifiersEXT`. Chromium's
+`OzoneImageGLTexturesHolder::GetBinding` only picked
+`GL_TEXTURE_EXTERNAL_OES` when the SharedImageFormat carried
+`PrefersExternalSampler` — which is set for the generic Linux multi-plane
+case but **not** for V4L2 producers that arrive via the standard ozone
+pixmap path. The frame then took the `GL_TEXTURE_2D` branch, ANGLE's
+`validationES.cpp:4894` rejected the YUV EGLImage on a non-EXTERNAL_OES
+target, the import failed, and the renderer fell back to the
+NV12→AR24 software conversion (~131 % CPU baseline).
+
+Patch closes the gap: also pick `EXTERNAL_OES` when the EGL driver
+advertises the pixmap's modifier as `external_only` (cached per
+`(fourcc, modifier)` tuple via a function-local
+`base::flat_map`+`base::Lock`, so the EGL round-trip stays off the
+per-frame hot path). Adds a single static helper
+`NativePixmapEGLBinding::ModifierRequiresExternalOES`. ~+90 lines, zero
+deletions, no shader changes (Skia Ganesh already handles
+`GL_TEXTURE_EXTERNAL_OES` natively via `GrGLTextureInfo.fTarget`).
+
+### Validation on ohm (RK3566 PineTab2 / hantro mainline 6.19.10)
+- `bbb_1080p30_h264.mp4` plays clean, no garble, no decoder error
+- Steady-state **34.7 % combined CPU** during 1080p30 H.264 (browser 12 +
+  GPU 9 + net 6 + render 6 + audio 1) vs v3 baseline ~131 % — **~3.8×
+  reduction**. Risk-1 (ANGLE+EXTERNAL_OES sampling regression on Skia
+  Ganesh / panfrost) **cleared**.
+- `V4L2VideoDecoder()` constructor + `Using a stateless API for profile:
+  h264 main and fourcc: S264` confirmed in log; 6 CAPTURE buffers
+  V4L2_MEMORY_MMAP, NV12 output. 19 live dmabuf fds in GPU process
+  during steady playback — healthy V4L2 rotation + compositor depth, not
+  a leak.
+
+### KWin 6.6.4 GL_ALPHA bug — separate, preexisting, blocks long playback
+Across BOTH v3 (no patch 4) and v4 (with patch 4) chromium-fourier
+builds, mid-playback the renderer + GPU process both park in
+`futex_do_wait`, `<video>` element keeps its ⏸ icon, currentTime
+advances on the audio clock, and audio outputs static (last ALSA buffer
+recycled) then silence. No D-state, no v4l2/vb2/dma_fence wchan, no
+error in `chrome-v[34].log`.
+
+`journalctl` for `kwin_wayland`:
+```
+GL_INVALID_VALUE in glTexImage2D(internalFormat=GL_ALPHA)
+GL_INVALID_OPERATION in glTexSubImage2D(invalid texture level 0) × N
+```
+First occurrence on this box: **2026-03-06**. KWin is requesting an
+internal format that doesn't exist in modern GLES (`GL_ALPHA` is GLES1.x
+legacy, not valid for `glTexImage2D` with GLES3 contexts). The
+allocation fails, then every `glTexSubImage2D` to that texture errors at
+level 0; KWin keeps retrying the same broken upload every frame, never
+recovers. The frame-callback ack to wayland clients stalls → chrome's
+renderer parks waiting for the present-feedback that never lands.
+
+Patch 4 looks "guilty" only because of timing: with NV12 zero-copy, the
+renderer is fast enough to actually post a v4l2-backed dmabuf within the
+window where KWin's broken path runs; v3 was slow enough (NV12→AR24
+software conversion) that the bug surfaced 5–10× later. **Triangulation:**
+chrome v4 stalls + chrome v3 stalls + VLC `cannot convert decoder/filter
+output` + mpv `could not initialize video chain` — every wayland video
+client hits it; ffmpeg `-hwaccel v4l2request -f null` plays through
+clean (decode path is healthy, the wall is the compositor's GL backend).
+
+### Decoder-stack sanity post-reboot (2026-04-28 ~13:30)
+After a reboot the V4L2 device numbering shuffled:
+- `/dev/video0` = `rockchip,rk3568-vpu-dec` (hantro DEC, was video1)
+- `/dev/video1` = `rockchip,rk3568-vepu-enc` (hantro ENC)
+- `/dev/video2` = `rockchip-rga`
+- `/dev/media0` = controller for DEC, `/dev/media1` = controller for ENC
+
+Anything that hardcoded `/dev/video1` for decode now talks to the
+encoder. Chrome and ffmpeg both handle this transparently (they enumerate
+via media-ctl); mpv's `--hwdec=v4l2request` returns `Could not create
+device` post-shuffle — separate mpv-side bug, not ours.
+
+### Outstanding (revised, supersedes earlier list)
+1. **Patch 4 lands publicly:** bump PKGBUILD `source=` and `prepare()`,
+   commit + tag a `chromium-fourier-149-r4` release on
+   `git@github.com:marfrit/chromium-fourier`.
+2. **KWin pivot** — see `KWIN_PIVOT.md` (separate doc) for the plan to
+   identify and patch the `glTexImage2D(GL_ALPHA)` site, since ohm is the
+   only board on hand and every wayland video client is affected.
+3. **Replication on ampere** (RK3588, panthor + rkvdec2 + hantro
+   multiplanar) — needs ampere woken; currently DOWN.
+4. **firefox-fourier 150 build** — `firefox-fourier-150.0.1-1-aarch64.pkg.tar.zst`
+   is built (95 MB on workstation:/tmp/, sha256 acbf1870…), pending
+   fresnel power-on for V4L2 stateless validation on RK3399.

arch/chromium-fourier/PKGBUILD

@@ -4,19 +4,20 @@
 # (RK3566 hantro / RK3588 VDPU381) on **mainline** kernel + Wayland +
 # panfrost / panthor — no vendor MPP, no Mali blob, no panfork, no
 # 5.10 BSP kernel. Fills the niche that 7Ji's chromium-mpp explicitly
-# does not (it forces BSP + X11 + vendor stack); see
-# /home/mfritsche/src/marfrit-packages/arch/chromium-fourier/STUDY.md
-# for the full rationale.
+# does not (it forces BSP + X11 + vendor stack); see STUDY.md and
+# NEXT.md alongside this PKGBUILD for the full rationale and the
+# validation log on PineTab2 (RK3566).
 #
-# Build host: chromium-builder LXD container on boltzmann (8 cores,
-# 28 GB RAM cap, 824 GB NVMe). 6-10 h initial build.
+# Multi-arch: builds natively on x86_64 and aarch64. The x86_64 path
+# is primarily a development / CI host; the runtime target audience is
+# aarch64. The two patches are architecture-independent.
 
 pkgname=chromium-fourier
 pkgver=147.0.7727.116
-pkgrel=1
+pkgrel=2
 epoch=1
 pkgdesc='Chromium with V4L2VDA HW video decode unlocked for mainline Linux Wayland on Rockchip'
-arch=('aarch64')
+arch=('aarch64' 'x86_64')
 url='https://www.chromium.org/Home'
 license=('BSD-3-Clause')
 depends=(
@@ -68,45 +69,84 @@ optdepends=(
 provides=(chromium)
 conflicts=(chromium)
 options=('!lto' '!strip')
-# NB: the chromium tarball is fetched lazily by build.sh on the build
-# host (5.7 GB compressed, doesn't belong in this git repo). The
-# PKGBUILD assumes /build/chromium/src is already populated.
+
+# Canonical chromium release tarball (5.7 GB compressed). Versions track
+# the chromium release schedule (see https://chromiumdash.appspot.com).
+# When bumping pkgver the patches may need their hunk line numbers
+# refreshed against the new tree — they are written against
+# media/base/media_switches.cc and ui/ozone/platform/wayland/gpu/wayland_surface_factory.cc
+# which both move around between minor releases.
 source=(
-  'patches/chromeos-pipeline-bypass.patch'
+  "https://commondatastorage.googleapis.com/chromium-browser-official/chromium-${pkgver}.tar.xz"
+  'patches/enable-v4l2-decoder-default.patch'
+  'patches/wayland-allow-direct-egl-gles2.patch'
+  'patches/nv12-external-oes-on-modifier-external-only.patch'
 )
 sha256sums=(
   'SKIP'
+  'SKIP'
+  'SKIP'
+  'SKIP'
 )
 
-# NB: this PKGBUILD is currently development-shaped — it operates on a
-# pre-extracted chromium tree at /build/chromium/src rather than letting
-# makepkg fetch + unpack the 5.7 GB tarball into ${srcdir}. Once the
-# patches stabilise we'll switch to the canonical pattern with
-# source=(${url}/chromium-${pkgver}.tar.xz ...) so the standard makepkg
-# flow + CI fermi-style pipeline work end-to-end.
-
 prepare() {
-  cd /build/chromium/src
+  cd "${srcdir}/chromium-${pkgver}"
 
-  # Fourier-local: bypass the chromeos pipeline so VaapiVideoDecoder /
-  # V4L2VDA is reachable on Linux non-ChromeOS. The 7Ji-style gn args
-  # (use_v4l2_codec / use_v4lplugin / use_linux_v4l2_only) may be
-  # sufficient on their own; this patch is the fallback if they aren't.
-  patch -Np1 -i "${srcdir}/patches/chromeos-pipeline-bypass.patch" || true
+  # Fourier patch 1/2: flip kAcceleratedVideoDecodeLinux's default to
+  # enabled when USE_V4L2_CODEC is the build's HW decode path. Without
+  # this the runtime master gate stays off on USE_V4L2_CODEC-only builds
+  # and chrome silently falls back to ffmpeg software decode. See the
+  # patch header for the validation log on RK3566 hantro.
+  patch -Np1 -i "${srcdir}/patches/enable-v4l2-decoder-default.patch"
+
+  # Fourier patch 2/3: re-allow the direct EGL/GLES2 path in the Wayland
+  # ozone surface factory so panfrost's EGL_EXT_image_dma_buf_import
+  # surfaces to chrome's GL display, lighting up the NV12 zero-copy
+  # native-pixmap pipeline. The launcher defaults to ANGLE (DCHECK in
+  # gl_context_egl.cc:241 fires on direct EGL with non-official builds);
+  # this patch keeps the direct path available for users who flip
+  # is_official_build=true and want the lower-CPU pipeline.
+  patch -Np1 -i "${srcdir}/patches/wayland-allow-direct-egl-gles2.patch"
+
+  # Fourier patch 3/3: pick GL_TEXTURE_EXTERNAL_OES for NV12 dmabufs
+  # whose DRM modifier is advertised external_only by the EGL driver.
+  # On panfrost / panthor every NV12 modifier (LINEAR + AFBC + AFRC) is
+  # external_only; chromium's default OzoneImageGLTexturesHolder picked
+  # GL_TEXTURE_2D and ANGLE then rejected the YUV EGLImage on a
+  # non-EXTERNAL_OES target, forcing the NV12->AR24 software conversion
+  # fallback. This closes that gap and enables the actual zero-copy
+  # path. Validated on ohm (RK3566 hantro): 1080p30 H.264 drops from
+  # ~131% combined CPU to ~34.7% (~3.8x). See patches/0004 for context.
+  patch -Np1 -i "${srcdir}/patches/nv12-external-oes-on-modifier-external-only.patch"
 
   # Use system node, system java
-  rm -f third_party/node/linux/node-linux-arm64/bin/node
-  mkdir -p third_party/node/linux/node-linux-arm64/bin
-  ln -sf /usr/bin/node third_party/node/linux/node-linux-arm64/bin/
+  case "$CARCH" in
+    aarch64) _node_dir=node-linux-arm64 ;;
+    x86_64)  _node_dir=node-linux-x64 ;;
+  esac
+  rm -f "third_party/node/linux/${_node_dir}/bin/node"
+  mkdir -p "third_party/node/linux/${_node_dir}/bin"
+  ln -sf /usr/bin/node "third_party/node/linux/${_node_dir}/bin/"
   ln -sf /usr/bin/java third_party/jdk/current/bin/ 2>/dev/null || true
 }
 
 build() {
-  cd /build/chromium/src
+  cd "${srcdir}/chromium-${pkgver}"
+
+  case "$CARCH" in
+    aarch64) _target_cpu="arm64" ;;
+    x86_64)  _target_cpu="x64"   ;;
+  esac
 
   local _flags=(
-    'is_official_build=true'
+    "target_cpu=\"${_target_cpu}\""
+
+    'is_official_build=false'
     'is_debug=false'
+    # dcheck_always_on defaults to !is_official_build (true here);
+    # explicitly off so the direct EGL/GLES2 path doesn't FATAL on
+    # gl_context_egl.cc:241's DCHECK(!global_texture_share_group_).
+    'dcheck_always_on=false'
     'symbol_level=0'
     'is_cfi=false'
     'treat_warnings_as_errors=false'
@@ -123,7 +163,7 @@ build() {
     'use_v4l2_codec=true'
     'use_v4lplugin=true'
     'use_linux_v4l2_only=true'
-    'use_vaapi=true'
+    'use_vaapi=false'
 
     # Codec branding for proprietary codec support (H.264 etc.)
     'ffmpeg_branding="Chrome"'
@@ -136,20 +176,38 @@ build() {
   )
 
   gn gen out/Default --args="${_flags[*]}"
-  autoninja -C out/Default chrome chromedriver chrome_sandbox
+  ninja -C out/Default chrome chrome_crashpad_handler
 }
 
 package() {
-  cd /build/chromium/src
+  cd "${srcdir}/chromium-${pkgver}"
 
   install -Dm755 out/Default/chrome "${pkgdir}/usr/lib/chromium/chromium"
-  install -Dm4755 out/Default/chrome_sandbox "${pkgdir}/usr/lib/chromium/chrome-sandbox"
-  install -Dm755 out/Default/chromedriver "${pkgdir}/usr/bin/chromedriver"
+  install -Dm755 out/Default/chrome_crashpad_handler \
+    "${pkgdir}/usr/lib/chromium/chrome_crashpad_handler"
+  [ -f out/Default/chrome_sandbox ] && install -Dm4755 out/Default/chrome_sandbox \
+    "${pkgdir}/usr/lib/chromium/chrome-sandbox"
+  [ -f out/Default/chromedriver ] && install -Dm755 out/Default/chromedriver \
+    "${pkgdir}/usr/bin/chromedriver"
+
+  # Bundled GL/Vulkan runtime — chrome dlopens these from its own dir,
+  # not /usr/lib/. Without them GL init fails and chrome falls back to
+  # software compositing.
+  for so in libEGL.so libGLESv2.so libvk_swiftshader.so libvulkan.so.1; do
+    [ -f "out/Default/$so" ] && install -Dm755 "out/Default/$so" \
+      "${pkgdir}/usr/lib/chromium/$so"
+  done
+  # ANGLE and SwiftShader ICD config files
+  for icd in out/Default/*_icd.json; do
+    [ -f "$icd" ] && install -Dm644 "$icd" \
+      "${pkgdir}/usr/lib/chromium/$(basename "$icd")"
+  done
 
   # Resources / locales / pak files
   for f in chrome_100_percent.pak chrome_200_percent.pak resources.pak \
-           v8_context_snapshot.bin icudtl.dat headless_lib_data.pak \
-           headless_lib_strings.pak; do
+           v8_context_snapshot.bin snapshot_blob.bin icudtl.dat \
+           headless_lib_data.pak headless_lib_strings.pak \
+           headless_command_resources.pak; do
     [ -f "out/Default/$f" ] && install -Dm644 "out/Default/$f" \
       "${pkgdir}/usr/lib/chromium/$f"
   done
@@ -160,7 +218,48 @@ package() {
     cp -r out/Default/locales/* "${pkgdir}/usr/lib/chromium/locales/"
   fi
 
-  # Top-level launcher
+  # Launcher shim — defaults to ANGLE→GLES on Wayland with Vulkan
+  # disabled. Vulkan is off by default because:
+  #   - panvk on RK3566 (Mali-G52 Bifrost) returns
+  #     VK_ERROR_INCOMPATIBLE_DRIVER on chromium's probe and breaks
+  #     V4L2 dispatch downstream (chrome falls back to FFmpeg software);
+  #   - panthor on RK3588 (Mali-G610 Valhall) is more functional but
+  #     not yet validated end-to-end against this build.
+  #
+  # User overrides for development on other Rockchips:
+  #   --enable-features=Vulkan         enable Vulkan (panthor / others)
+  #   --use-vulkan=native|swiftshader  pick the Vulkan backend
+  #   --disable-features=Vulkan        explicit re-disable
+  # Any of those on the command line short-circuits the launcher's
+  # default disable, so the user's intent always wins.
   install -dm755 "${pkgdir}/usr/bin"
-  ln -s /usr/lib/chromium/chromium "${pkgdir}/usr/bin/chromium"
+  cat > "${pkgdir}/usr/bin/chromium" <<'LAUNCHER'
+#!/bin/bash
+# chromium-fourier launcher — V4L2 HW decode + Wayland + ANGLE
+# Vulkan disabled by default; pass --enable-features=Vulkan or
+# --use-vulkan=native to opt in (e.g. RK3588 panthor work).
+
+USER_HANDLES_VULKAN=0
+for arg in "$@"; do
+  case "$arg" in
+    --use-vulkan*|--enable-features=*Vulkan*|--disable-features=*Vulkan*|--use-angle=vulkan*)
+      USER_HANDLES_VULKAN=1
+      break
+      ;;
+  esac
+done
+
+vulkan_default=()
+if [ "$USER_HANDLES_VULKAN" = 0 ]; then
+  vulkan_default=(--disable-features=Vulkan)
+fi
+
+exec /usr/lib/chromium/chromium \
+  --ozone-platform=wayland \
+  --use-gl=angle --use-angle=gles \
+  --enable-features=AcceleratedVideoDecoder \
+  "${vulkan_default[@]}" \
+  "$@"
+LAUNCHER
+  chmod 0755 "${pkgdir}/usr/bin/chromium"
 }

arch/chromium-fourier/patches/chromeos-pipeline-bypass.patch

@@ -1,22 +0,0 @@
-From: Markus Fritsche <mfritsche@reauktion.de>
-Subject: media/gpu: skip chromeos VideoDecoderPipeline on non-ChromeOS Linux
-
-Placeholder. The patch will be developed against the actual chromium
-147.0.7727.116 source tree on chromium-builder@boltzmann once the
-tarball is extracted and we can read the exact code paths in:
-
-  - media/gpu/chromeos/video_decoder_pipeline.cc (PickDecoderOutputFormat,
-    Initialize, ImageProcessor setup)
-  - media/gpu/vaapi/vaapi_video_decoder.cc (ApplyResolutionChangeWithScreenSizes,
-    line ~1219 where "failed Initialize()ing the frame pool" fires)
-
-The 7Ji-style gn args (`use_v4l2_codec=true use_v4lplugin=true
-use_linux_v4l2_only=true`) MAY be sufficient by themselves to route
-decode through the legacy V4L2VDA path entirely, bypassing the
-chromeos pipeline without source patches. First build will tell us;
-this file is the placeholder for the patch we'll need if it isn't.
-
-# Empty no-op patch. patch -p1 < this won't change anything.
-diff --git a/PLACEHOLDER b/PLACEHOLDER
-new file mode 100644
-index 0000000..e69de29

arch/chromium-fourier/patches/compiler-rt-adjust-paths.patch

@@ -0,0 +1,38 @@
+From: 7Ji <7Ji@example.com> (originally), adapted for chromium-fourier
+Subject: Adjust compiler-rt library path layout for system clang on Arch
+ Linux ARM, where compiler-rt installs to lib/clang/N/lib/linux/ with
+ -aarch64 filename suffix instead of chromium's expected
+ lib/clang/N/lib/aarch64-unknown-linux-gnu/ layout.
+
+diff --git a/build/config/clang/BUILD.gn b/build/config/clang/BUILD.gn
+index d4de2e0cca0..57359c32121 100644
+--- a/build/config/clang/BUILD.gn
++++ b/build/config/clang/BUILD.gn
+@@ -130,12 +130,15 @@ template("clang_lib") {
+       } else if (is_linux || is_chromeos) {
+         if (current_cpu == "x64") {
+           _dir = "x86_64-unknown-linux-gnu"
++          _suffix = "-x86_64"
+         } else if (current_cpu == "x86") {
+           _dir = "i386-unknown-linux-gnu"
++          _suffix = "-i386"
+         } else if (current_cpu == "arm") {
+           _dir = "armv7-unknown-linux-gnueabihf"
+         } else if (current_cpu == "arm64") {
+           _dir = "aarch64-unknown-linux-gnu"
++          _suffix = "-aarch64"
+         } else {
+           assert(false)  # Unhandled cpu type
+         }
+@@ -166,6 +169,11 @@ template("clang_lib") {
+         assert(false)  # Unhandled target platform
+       }
+
++      # Bit of a hack to make this find builtins from compiler-rt >= 16
++      if (is_linux || is_chromeos) {
++        _dir = "linux"
++      }
++
+       _clang_lib_dir = "$clang_base_path/lib/clang/$clang_version/lib"
+       _lib_file = "${_prefix}clang_rt.${_libname}${_suffix}.${_ext}"
+       libs = [ "$_clang_lib_dir/$_dir/$_lib_file" ]

arch/chromium-fourier/patches/enable-v4l2-decoder-default.patch

@@ -0,0 +1,55 @@
+From: Markus Fritsche <mfritsche@reauktion.de>
+Subject: media: default kAcceleratedVideoDecodeLinux to enabled when
+ USE_V4L2_CODEC is the build's hardware decode path
+Date: 2026-04-26
+
+Background
+----------
+chromium-fourier targets mainline-Linux Wayland on Rockchip (RK3566 hantro,
+RK3588 VDPU381) where the only HW video decode path is V4L2 stateless
+(via the in-tree media/gpu/v4l2 stack). The build is configured with
+
+    use_vaapi      = false
+    use_v4l2_codec = true
+    use_v4lplugin  = true
+    use_linux_v4l2_only = true
+
+Without this patch, GPU-process V4L2 decode is compiled in but stays
+runtime-disabled by default. The runtime master gate
+`media::kAcceleratedVideoDecodeLinux` (the user-visible feature name is
+"AcceleratedVideoDecoder") is currently flipped to ENABLED_BY_DEFAULT only
+when `BUILDFLAG(USE_VAAPI)` is set. On a USE_V4L2_CODEC-only build the
+feature stays DISABLED_BY_DEFAULT, the linux gpu_mojo_media_client returns
+`VideoDecoderType::kUnknown`, and `<video>` falls all the way back to
+`media/filters/ffmpeg_video_decoder.cc` (software).
+
+We confirmed this by hand on the PineTab2 (RK3566 hantro): with
+`--enable-features=AcceleratedVideoDecoder` chrome correctly selects
+`V4L2VideoDecoder` for h264 main, opens /dev/video1 + /dev/media0,
+allocates 17 OUTPUT + 6 CAPTURE NV12 buffers, and runs SetExtCtrlsInit for
+H264. Without the runtime flag, none of that happens.
+
+Fix
+---
+Treat `USE_V4L2_CODEC` symmetrically with `USE_VAAPI` for the runtime
+default of the master gate. A user can still disable it via
+`--disable-features=AcceleratedVideoDecoder`.
+
+This does NOT touch the `kAcceleratedVideoDecodeLinuxGL` companion gate
+(already ENABLED_BY_DEFAULT) or any of the per-decoder selection logic in
+`media/mojo/services/gpu_mojo_media_client_linux.cc` -- that file already
+dispatches to the V4L2 decoder when `USE_V4L2_CODEC && !USE_VAAPI`, gated
+behind the master flag we are flipping here.
+
+diff --git a/media/base/media_switches.cc b/media/base/media_switches.cc
+--- a/media/base/media_switches.cc
++++ b/media/base/media_switches.cc
+@@ -749,7 +749,7 @@ BASE_FEATURE(kUnifiedAutoplay, base::FEATURE_ENABLED_BY_DEFAULT);
+ // on chromeos, but needs an experiment on linux.
+ BASE_FEATURE(kAcceleratedVideoDecodeLinux,
+              "AcceleratedVideoDecoder",
+-#if BUILDFLAG(USE_VAAPI)
++#if BUILDFLAG(USE_VAAPI) || BUILDFLAG(USE_V4L2_CODEC)
+              base::FEATURE_ENABLED_BY_DEFAULT);
+ #else
+              base::FEATURE_DISABLED_BY_DEFAULT);

arch/chromium-fourier/patches/nv12-external-oes-on-modifier-external-only.patch

@@ -0,0 +1,187 @@
+From: Markus Fritsche <mfritsche@reauktion.de>
+Subject: [PATCH] gpu/ozone: pick GL_TEXTURE_EXTERNAL_OES for NV12 dmabufs whose
+ DRM modifier is advertised external_only by the EGL driver
+Date: 2026-04-28
+
+Background
+----------
+On mainline-Linux Mali GPUs (mesa panfrost / panthor on Bifrost / Valhall)
+every NV12 modifier exposed by `eglQueryDmaBufModifiersEXT` is flagged
+`external_only` — DRM_FORMAT_MOD_LINEAR + ARM AFBC × 2 + ARM AFRC. Mesa's
+behavior is spec-correct: GLES sampling of multi-plane formats is
+defined only via `samplerExternalOES`, never `sampler2D`. The chromium
+NV12 import path at
+`gpu/command_buffer/service/shared_image/ozone_image_gl_textures_holder.cc`
+already chooses `GL_TEXTURE_EXTERNAL_OES` when the SharedImageFormat is
+flagged `PrefersExternalSampler` — but that flag is only set for the
+generic "multi-plane on Linux" case in
+`media/gpu/chromeos/mailbox_video_frame_converter.cc`. Frames that
+arrive with an `external_only`-flagged modifier from a producer that
+didn't set the flag (V4L2 hantro NV12 with AFBC/AFRC capture format on
+RK3588's rkvdec2, future NativePixmap producers, etc.) hit the
+`GL_TEXTURE_2D` path; ANGLE's `validationES.cpp:4894` then rejects YUV
+EGLImages on non-EXTERNAL_OES targets, and the import fails.
+
+This patch closes the gap: the texture-target choice in
+`OzoneImageGLTexturesHolder::GetBinding` now consults the EGL driver's
+`external_only` annotation for the pixmap's actual modifier in addition
+to `format.PrefersExternalSampler()`. If either says "external sampler
+required", the target switches to `GL_TEXTURE_EXTERNAL_OES`. Skia
+Ganesh handles `GL_TEXTURE_EXTERNAL_OES` natively via
+`GrGLTextureInfo.fTarget`, so no shader changes are required. Same
+infrastructure chromium already uses for Android camera / decoder
+dmabufs, retargeted at the Linux ozone layer.
+
+Result is cached per `(fourcc, modifier)` tuple via a function-local
+static `base::flat_map`, so the EGL query is not on the per-frame hot
+path — once per unique format+modifier combination, after which the
+runtime cost is a hash lookup behind a base::Lock.
+
+Bug crbug.com/1498703 is the closest existing tracker; framing this
+upstream as "make Linux NV12 import path consistent with the
+ChromeOS PrefersExternalSampler default" is the right angle.
+
+diff --git a/gpu/command_buffer/service/shared_image/ozone_image_gl_textures_holder.cc b/gpu/command_buffer/service/shared_image/ozone_image_gl_textures_holder.cc
+index 525bdcb0dc..43b0723326 100644
+--- a/gpu/command_buffer/service/shared_image/ozone_image_gl_textures_holder.cc
++++ b/gpu/command_buffer/service/shared_image/ozone_image_gl_textures_holder.cc
+@@ -16,6 +16,7 @@
+ #include "ui/gl/gl_bindings.h"
+ #include "ui/gl/scoped_binders.h"
+ #include "ui/ozone/public/gl_ozone.h"
++#include "ui/ozone/common/native_pixmap_egl_binding.h"
+ #include "ui/ozone/public/native_pixmap_gl_binding.h"
+ #include "ui/ozone/public/ozone_platform.h"
+ #include "ui/ozone/public/surface_factory_ozone.h"
+@@ -82,7 +83,14 @@ std::unique_ptr<ui::NativePixmapGLBinding> GetBinding(
+   // being multiplanar (if using per-plane sampling of a multiplanar texture,
+   // the buffer format passed in here must be the single-planar format of the
+   // plane).
+-  if (format.PrefersExternalSampler()) {
++  // chromium-fourier: also pick GL_TEXTURE_EXTERNAL_OES whenever the
++  // pixmap's DRM modifier is advertised external_only by the EGL
++  // driver. Mesa panfrost / panthor mark every NV12 modifier
++  // external_only — the PrefersExternalSampler flag alone misses
++  // the AFBC / AFRC tiled paths.
++  if (format.PrefersExternalSampler() ||
++      ui::NativePixmapEGLBinding::ModifierRequiresExternalOES(
++          pixmap.get(), plane_format)) {
+     target = GL_TEXTURE_EXTERNAL_OES;
+   } else {
+     target = GL_TEXTURE_2D;
+diff --git a/ui/ozone/common/native_pixmap_egl_binding.cc b/ui/ozone/common/native_pixmap_egl_binding.cc
+index 31877f4459..6855c1093e 100644
+--- a/ui/ozone/common/native_pixmap_egl_binding.cc
++++ b/ui/ozone/common/native_pixmap_egl_binding.cc
+@@ -6,10 +6,13 @@
+ 
+ #include <array>
+ 
++#include "base/containers/flat_map.h"
+ #include "base/logging.h"
+ #include "base/memory/scoped_refptr.h"
++#include "base/no_destructor.h"
+ #include "base/notreached.h"
+ #include "base/numerics/safe_conversions.h"
++#include "base/synchronization/lock.h"
+ #include "ui/gfx/linux/drm_util_linux.h"
+ #include "ui/gl/gl_bindings.h"
+ #include "ui/gl/gl_surface_egl.h"
+@@ -56,6 +59,75 @@ bool NativePixmapEGLBinding::IsSharedImageFormatSupported(
+     viz::SharedImageFormat format) {
+   return GetFourCCFormatFromSharedImageFormat(format) != DRM_FORMAT_INVALID;
+ }
++// static
++bool NativePixmapEGLBinding::ModifierRequiresExternalOES(
++    const gfx::NativePixmap* pixmap,
++    viz::SharedImageFormat format) {
++  // chromium-fourier: query the EGL driver for the (fourcc, modifier)
++  // tuple's external_only flag. Cache results — eglQueryDmaBufModifiersEXT
++  // is a synchronous round-trip into the driver and we want it off the
++  // per-frame hot path. The cache lives for the lifetime of the GPU
++  // process (modifier tables don't change after EGL init).
++  if (!pixmap) {
++    return false;
++  }
++  const uint64_t modifier = pixmap->GetFormatModifier();
++  if (modifier == gfx::NativePixmapHandle::kNoModifier) {
++    // Implicit linear — same answer the driver would give for the
++    // matching LINEAR entry, but cheaper not to query.
++    return false;
++  }
++  const uint32_t fourcc = GetFourCCFormatFromSharedImageFormat(format);
++  if (fourcc == DRM_FORMAT_INVALID) {
++    return false;
++  }
++
++  using Key = std::pair<uint32_t, uint64_t>;
++  static base::NoDestructor<base::Lock> cache_lock;
++  static base::NoDestructor<base::flat_map<Key, bool>> cache;
++  {
++    base::AutoLock lock(*cache_lock);
++    auto it = cache->find({fourcc, modifier});
++    if (it != cache->end()) {
++      return it->second;
++    }
++  }
++
++  bool external_only = false;
++  do {
++    auto* display = gl::GLSurfaceEGL::GetGLDisplayEGL();
++    if (!display || !display->ext->b_EGL_EXT_image_dma_buf_import_modifiers) {
++      break;
++    }
++    EGLDisplay egl_display = display->GetDisplay();
++    EGLint num_modifiers = 0;
++    if (!eglQueryDmaBufModifiersEXT(egl_display, fourcc, 0, nullptr, nullptr,
++                                    &num_modifiers) ||
++        num_modifiers <= 0) {
++      break;
++    }
++    std::vector<EGLuint64KHR> modifiers(num_modifiers);
++    std::vector<EGLBoolean> ext_only(num_modifiers);
++    if (!eglQueryDmaBufModifiersEXT(egl_display, fourcc, num_modifiers,
++                                    modifiers.data(), ext_only.data(),
++                                    &num_modifiers)) {
++      break;
++    }
++    for (EGLint i = 0; i < num_modifiers; ++i) {
++      if (modifiers[i] == modifier) {
++        external_only = (ext_only[i] == EGL_TRUE);
++        break;
++      }
++    }
++  } while (0);
++
++  {
++    base::AutoLock lock(*cache_lock);
++    cache->insert_or_assign({fourcc, modifier}, external_only);
++  }
++  return external_only;
++}
++
+ 
+ // static
+ std::unique_ptr<NativePixmapGLBinding> NativePixmapEGLBinding::Create(
+diff --git a/ui/ozone/common/native_pixmap_egl_binding.h b/ui/ozone/common/native_pixmap_egl_binding.h
+index 61fb0de77f..ad3ac9ced5 100644
+--- a/ui/ozone/common/native_pixmap_egl_binding.h
++++ b/ui/ozone/common/native_pixmap_egl_binding.h
+@@ -27,6 +27,17 @@ class NativePixmapEGLBinding : public NativePixmapGLBinding {
+ 
+   static bool IsSharedImageFormatSupported(viz::SharedImageFormat format);
+ 
++  // chromium-fourier: returns true when |pixmap|'s DRM format modifier
++  // is advertised by the EGL driver as `external_only` for the given
++  // SharedImage format. Used at SharedImage creation time to override
++  // the default GL_TEXTURE_2D target to GL_TEXTURE_EXTERNAL_OES so that
++  // mesa panfrost / panthor NV12 dmabufs (always external_only) import
++  // cleanly via glEGLImageTargetTexture2DOES + samplerExternalOES.
++  // Result is cached per (fourcc, modifier) tuple — the underlying
++  // eglQueryDmaBufModifiersEXT call is not on the per-frame hot path.
++  static bool ModifierRequiresExternalOES(const gfx::NativePixmap* pixmap,
++                                          viz::SharedImageFormat format);
++
+   // Create an EGLImage from a given NativePixmap and plane and bind
+   // |texture_id| to |target| followed by binding the image to |target|. The
+   // color space is for the external sampler: When we sample the YUV buffer as

arch/chromium-fourier/patches/wayland-allow-direct-egl-gles2.patch

@@ -0,0 +1,57 @@
+From: Markus Fritsche <mfritsche@reauktion.de>
+Subject: ozone/wayland: re-allow direct EGL/GLES2 path (no ANGLE shim)
+Date: 2026-04-26
+
+Background
+----------
+On Wayland-only ozone builds the surface factory currently advertises only
+ANGLE-mediated GL implementations (`kOpenGL`, `kOpenGLES`, `kSwiftShader`,
+`kVulkan`). Anything driving `--use-gl=egl` (the historical
+`kGLImplementationEGLGLES2`) is rejected at startup with
+
+    Requested GL implementation (gl=egl-gles2,angle=none) not found in
+    allowed implementations: [(gl=egl-angle,angle=opengl|opengles|...)]
+
+The downstream switch already handles `kGLImplementationEGLGLES2` in
+`GetGLOzone`, so the dispatcher is wired -- it's the *advertisement* that
+got tightened.
+
+The cost of that tightening on RK3566 hantro / panfrost is real: with
+ANGLE in the path, chrome's display-side EGL is ANGLE's own EGL. ANGLE's
+GLES backend on Linux does not propagate
+`EGL_EXT_image_dma_buf_import` through to the chrome GL display, so
+`gpu_feature_info.supports_nv12_gl_native_pixmap` ends up false. That in
+turn forces the V4L2-decoded NV12 frames through the
+NV12-to-AR24 VPP conversion path before they hit the compositor, costing
+~85 % CPU at 1080p30 even though the hantro VPU is doing the actual
+decode for free.
+
+Allowing the direct EGL/GLES2 path back means chrome's EGL is panfrost's
+EGL (via mesa), which exposes the dmabuf-import extensions natively, and
+the zero-copy NV12 native pixmap path lights up.
+
+Fix
+---
+Add `kGLImplementationEGLGLES2` to the head of the allowed list; ANGLE
+remains the default fallback and is still selected when the user passes
+`--use-gl=angle ...`. The position is deliberate: on a USE_V4L2_CODEC
+hardware-decode build the user almost always wants the dmabuf-capable
+direct path; ANGLE is still there for browsers that need its conformance
+fixups.
+
+This does not affect non-Wayland ozone backends.
+
+diff --git a/ui/ozone/platform/wayland/gpu/wayland_surface_factory.cc b/ui/ozone/platform/wayland/gpu/wayland_surface_factory.cc
+--- a/ui/ozone/platform/wayland/gpu/wayland_surface_factory.cc
++++ b/ui/ozone/platform/wayland/gpu/wayland_surface_factory.cc
+@@ -223,6 +223,10 @@ std::vector<gl::GLImplementationParts>
+ WaylandSurfaceFactory::GetAllowedGLImplementations() {
+   std::vector<gl::GLImplementationParts> impls;
+   if (egl_implementation_) {
++    // chromium-fourier: keep the direct EGL/GLES2 path available so
++    // panfrost's EGL_EXT_image_dma_buf_import surfaces to chrome's GL
++    // display layer. See patch header for rationale.
++    impls.emplace_back(gl::GLImplementationParts(gl::kGLImplementationEGLGLES2));
+     impls.emplace_back(gl::ANGLEImplementation::kOpenGL);
+     impls.emplace_back(gl::ANGLEImplementation::kOpenGLES);
+     impls.emplace_back(gl::ANGLEImplementation::kSwiftShader);