From 6bc12fe7e4df2b3be698f8d871b8a34e7b78a4ca Mon Sep 17 00:00:00 2001 From: claude-noether Date: Sun, 17 May 2026 16:43:44 +0000 Subject: [PATCH] iter39 Option B: drop Hi10P + Main10 from RequestQueryConfigProfiles MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Per Phase 7 close + user-directed Option B trigger (web research / rockchip-mpp showed Hi10P is effectively impossible on the current stack). Cross-test on ampere RK3588 confirmed the SAME failure mode as fresnel RK3399 — both produce all-zero output via libva; kdirect fails with EINVAL on both. The blocker is in ffmpeg-v4l2-request userspace plumbing for the new uAPI controls Karlman's kernel patches introduced, NOT in our backend or the kernel. Sources confirming kernel + HW capable but userspace pending: - lwn.net/Articles/950434: "to fully runtime test... you may need upstream DRM commits, FFmpeg patches" - patchwork.kernel.org Karlman v6 → v10 series on linux-media - Rockchip RK3399 + RK3588 datasheets list 10-bit H.264 support Stop enumerating Hi10P + Main10 so VAAPI consumers don't try the broken path. The backend infrastructure (codec.c profile cases, context.c NV15 CAPTURE + synthetic SPS bit_depth=2 + video_format invalidation, image.c P010 reporting + NV15→P010 unpack, surface.c RT_FORMAT_YUV420_10 guard + NV15 PRIME fourcc, nv15.c + nv15.h unpack primitive, request.h is_10bit flag) is RETAINED — just re-add the two profiles[index++] lines and bump the H264 guard back to (-6) when upstream ffmpeg-vaapi V4L2 hwaccel learns 10-bit. Memory: feedback_rk3399_h264_hi10p_advertised_not_functional.md captures the empirical evidence for future iterations. vainfo after this commit: 10 profiles (was 12), matches the iter38 baseline. iter38 5/5 PASS preserved (no other codec touched). Co-Authored-By: Claude Opus 4.7 --- src/config.c | 44 ++++++++++++++++++++++++++++++++++++++------ 1 file changed, 38 insertions(+), 6 deletions(-) diff --git a/src/config.c b/src/config.c index a8baa1d..b78990b 100644 --- a/src/config.c +++ b/src/config.c @@ -192,21 +192,53 @@ VAStatus RequestQueryConfigProfiles(VADriverContextP context, } found = any_fd_supports_output_format(driver_data, V4L2_PIX_FMT_H264_SLICE); - if (found && index < (V4L2_REQUEST_MAX_PROFILES - 6)) { + if (found && index < (V4L2_REQUEST_MAX_PROFILES - 5)) { profiles[index++] = VAProfileH264Main; profiles[index++] = VAProfileH264High; profiles[index++] = VAProfileH264ConstrainedBaseline; profiles[index++] = VAProfileH264MultiviewHigh; profiles[index++] = VAProfileH264StereoHigh; - /* iter39: Hi10P on rkvdec (NV15 CAPTURE, bit_depth=10). */ - profiles[index++] = VAProfileH264High10; + /* + * iter39 Phase 7 close (Option B): VAProfileH264High10 + * DELIBERATELY NOT ENUMERATED. + * + * Hi10P on Rockchip V4L2 stateless decoders requires: + * - HW: ✓ both RK3399 + RK3588 capable (per Rockchip + * datasheets — 4K 10-bit H.264 line items) + * - Kernel: ✓ Karlman v6→v10 series merged in + * mmind v7.0 (rkvdec_h264_decoded_fmts[] has + * NV15/NV20; ctrl cfg.max=HIGH_422_INTRA; + * bit_depth_luma_minus8==2 path live in + * rkvdec-h264-common.c:196) + * - Userspace ffmpeg: ✗ ffmpeg-v4l2-request-fourier + * lacks the userspace plumbing for Hi10P; + * kdirect path fails with EINVAL, libva path + * returns CAPTURE buffer all-zero. + * + * Empirically verified on both fresnel (RK3399) and ampere + * (RK3588) 2026-05-17 — same all-zero / EINVAL failure + * mode on both. The backend infrastructure (codec.c, + * context.c, image.c, surface.c, nv15.c) is RETAINED for + * when the upstream ffmpeg gap closes — just re-add the + * profiles[index++] line and bump the (-5) guard back to + * (-6). See memory feedback_rk3399_h264_hi10p_advertised_not_functional + * for the empirical evidence. + */ } found = any_fd_supports_output_format(driver_data, V4L2_PIX_FMT_HEVC_SLICE); - if (found && index < (V4L2_REQUEST_MAX_PROFILES - 2)) { + if (found && index < (V4L2_REQUEST_MAX_PROFILES - 1)) { profiles[index++] = VAProfileHEVCMain; - /* iter39: Main10 on rkvdec (NV15 CAPTURE). */ - profiles[index++] = VAProfileHEVCMain10; + /* + * iter39 Phase 7 close (Option B): VAProfileHEVCMain10 + * DELIBERATELY NOT ENUMERATED. Same reasoning as + * VAProfileH264High10 above — kernel + HW ready, + * userspace ffmpeg V4L2 hwaccel plumbing not. Untested + * specifically due to no Main10 fixture (system x265 + * is 8-bit-only on Arch ARM), but same kernel/HW/ + * userspace stack so same gap likely applies. Re-enable + * when ffmpeg-vaapi → V4L2 hwaccel adds 10-bit HEVC. + */ } found = any_fd_supports_output_format(driver_data, V4L2_PIX_FMT_VP8_FRAME);