From patchwork Thu Dec 2 01:43:16 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Wenbin Chen X-Patchwork-Id: 31872 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a6b:d206:0:0:0:0:0 with SMTP id q6csp10108146iob; Wed, 1 Dec 2021 17:47:37 -0800 (PST) X-Google-Smtp-Source: ABdhPJwPr+qwQMRy1iQRavpDTqLQeiBYD+5Hy3U+biT9E/GMZWFEo9I6mkNFXVpaemX67yzhlQJ/ X-Received: by 2002:a17:906:128d:: with SMTP id k13mr11615022ejb.517.1638409657031; Wed, 01 Dec 2021 17:47:37 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1638409657; cv=none; d=google.com; s=arc-20160816; b=mY6/d/VChUGexSeOG2kPLJXOO1n/SXVTzpxY6+uFSrgNX6ivfl1kyfQ5pwE84hqi29 YENb1p6xXCIcHC1DJJWsrPm7ZMd3jK4PyMzndE94NuDcnzFoL3n+jqvU8adL8M1Kx0m3 R8T/i9KeSpHOro8ZWekBrLRrzNNECmfXtzpNhK2OxO3ciK+sjhf/KNMWcRXOOp3W0V0V zeEHJPuQnluIKRaQ7hy7SznMEkeOM6aLOTMfsaFtatnYKMYAJzsuaSJXNH0QdCrp6070 lXcH2euEZuQlNipj5DOFrnI1KCTWBiMcYiF/DZJ4qQo8D9fwLjOACWU742HZsrLSh0Vs wHZQ== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:reply-to:list-subscribe :list-help:list-post:list-archive:list-unsubscribe:list-id :precedence:subject:mime-version:message-id:date:to:from :delivered-to; bh=B9ZS4Q5STUEXuN7xPBnLS0OWXjWnqqhCX4dg195bg3Y=; b=mzwi/WAFB42/hiyUrKMO1Sm1ID/x8pSYx63803fhTfineHilkTQTTK/aBEVvGlgvkT 3tCKccs0fFxSRXs12TMebTruSGU3vKJhFUGEewn9iV8dk199gtojFjNIb5XLqqAxHr4Q wNjkxylikCterQ6H6+W5XEksfvLbiOnmDvCZJ/MLl3q3wFF81sKW6XKxV/Hyg5qJ/sxN ImhvgVFAPTj4kzYri5Tpujguq5FTUEYDyBz+X80Ffc0l7eu/zA+bWMJ2kFWz1eOHs4bP G6sQDLwbcKlyYw0lCt3bBVoWzjdANz/vRCsJRWm8iS21wdzAEsIgbhLzpeVuAaxWA06k IzRg== ARC-Authentication-Results: i=1; mx.google.com; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=intel.com Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id ho42si2852760ejc.50.2021.12.01.17.47.36; Wed, 01 Dec 2021 17:47:37 -0800 (PST) Received-SPF: pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) client-ip=79.124.17.100; Authentication-Results: mx.google.com; spf=pass (google.com: domain of ffmpeg-devel-bounces@ffmpeg.org designates 79.124.17.100 as permitted sender) smtp.mailfrom=ffmpeg-devel-bounces@ffmpeg.org; dmarc=fail (p=NONE sp=NONE dis=NONE) header.from=intel.com Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 887DF68AEBD; Thu, 2 Dec 2021 03:47:32 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 12E3668ABAC for ; Thu, 2 Dec 2021 03:47:25 +0200 (EET) X-IronPort-AV: E=McAfee;i="6200,9189,10185"; a="236414952" X-IronPort-AV: E=Sophos;i="5.87,280,1631602800"; d="scan'208";a="236414952" Received: from orsmga008.jf.intel.com ([10.7.209.65]) by orsmga102.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 01 Dec 2021 17:47:22 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.87,280,1631602800"; d="scan'208";a="512964508" Received: from chenwenbin-z390-aorus-ultra.sh.intel.com ([10.239.35.110]) by orsmga008.jf.intel.com with ESMTP; 01 Dec 2021 17:47:21 -0800 From: Wenbin Chen To: ffmpeg-devel@ffmpeg.org Date: Thu, 2 Dec 2021 09:43:16 +0800 Message-Id: <20211202014320.1066212-1-wenbin.chen@intel.com> X-Mailer: git-send-email 2.25.1 MIME-Version: 1.0 Subject: [FFmpeg-devel] [PATCH V5 1/5] hwcontext_vaapi: Use PRIME_2 memory type for modifiers. X-BeenThere: ffmpeg-devel@ffmpeg.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: FFmpeg development discussions and patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Reply-To: FFmpeg development discussions and patches Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: lJNdvd/RhJZr From: Bas Nieuwenhuizen This way we can pass explicit modifiers in. Sometimes the modifier matters for the number of memory planes that libva accepts, in particular when dealing with driver-compressed textures. Furthermore the driver might not actually be able to determine the implicit modifier if all the buffer-passing has used explicit modifier. All these issues should be resolved by passing in the modifier, and for that we switch to using the PRIME_2 memory type. Tested with experimental radeonsi patches for modifiers and kmsgrab. Also tested with radeonsi without the patches to double-check it works without PRIME_2 support. v2: Cache PRIME_2 support to avoid doing two calls every time on libva drivers that do not support it. v3: Remove prime2_vas usage. Signed-off-by: Bas Nieuwenhuizen --- libavutil/hwcontext_vaapi.c | 158 ++++++++++++++++++++++++++---------- 1 file changed, 114 insertions(+), 44 deletions(-) diff --git a/libavutil/hwcontext_vaapi.c b/libavutil/hwcontext_vaapi.c index 83e542876d..75acc851d6 100644 --- a/libavutil/hwcontext_vaapi.c +++ b/libavutil/hwcontext_vaapi.c @@ -79,6 +79,9 @@ typedef struct VAAPIFramesContext { unsigned int rt_format; // Whether vaDeriveImage works. int derive_works; + // Caches whether VA_SURFACE_ATTRIB_MEM_TYPE_DRM_PRIME_2 is unsupported for + // surface imports. + int prime_2_import_unsupported; } VAAPIFramesContext; typedef struct VAAPIMapping { @@ -1022,32 +1025,17 @@ static void vaapi_unmap_from_drm(AVHWFramesContext *dst_fc, static int vaapi_map_from_drm(AVHWFramesContext *src_fc, AVFrame *dst, const AVFrame *src, int flags) { + VAAPIFramesContext *src_vafc = src_fc->internal->priv; AVHWFramesContext *dst_fc = (AVHWFramesContext*)dst->hw_frames_ctx->data; AVVAAPIDeviceContext *dst_dev = dst_fc->device_ctx->hwctx; const AVDRMFrameDescriptor *desc; const VAAPIFormatDescriptor *format_desc; VASurfaceID surface_id; - VAStatus vas; + VAStatus vas = VA_STATUS_SUCCESS; + int use_prime2; uint32_t va_fourcc; - int err, i, j, k; - - unsigned long buffer_handle; - VASurfaceAttribExternalBuffers buffer_desc; - VASurfaceAttrib attrs[2] = { - { - .type = VASurfaceAttribMemoryType, - .flags = VA_SURFACE_ATTRIB_SETTABLE, - .value.type = VAGenericValueTypeInteger, - .value.value.i = VA_SURFACE_ATTRIB_MEM_TYPE_DRM_PRIME, - }, - { - .type = VASurfaceAttribExternalBufferDescriptor, - .flags = VA_SURFACE_ATTRIB_SETTABLE, - .value.type = VAGenericValueTypePointer, - .value.value.p = &buffer_desc, - } - }; + int err, i, j; desc = (AVDRMFrameDescriptor*)src->data[0]; @@ -1083,35 +1071,117 @@ static int vaapi_map_from_drm(AVHWFramesContext *src_fc, AVFrame *dst, format_desc = vaapi_format_from_fourcc(va_fourcc); av_assert0(format_desc); - buffer_handle = desc->objects[0].fd; - buffer_desc.pixel_format = va_fourcc; - buffer_desc.width = src_fc->width; - buffer_desc.height = src_fc->height; - buffer_desc.data_size = desc->objects[0].size; - buffer_desc.buffers = &buffer_handle; - buffer_desc.num_buffers = 1; - buffer_desc.flags = 0; - - k = 0; - for (i = 0; i < desc->nb_layers; i++) { - for (j = 0; j < desc->layers[i].nb_planes; j++) { - buffer_desc.pitches[k] = desc->layers[i].planes[j].pitch; - buffer_desc.offsets[k] = desc->layers[i].planes[j].offset; - ++k; + use_prime2 = !src_vafc->prime_2_import_unsupported && + desc->objects[0].format_modifier != DRM_FORMAT_MOD_INVALID; + if (use_prime2) { + VADRMPRIMESurfaceDescriptor prime_desc; + VASurfaceAttrib prime_attrs[2] = { + { + .type = VASurfaceAttribMemoryType, + .flags = VA_SURFACE_ATTRIB_SETTABLE, + .value.type = VAGenericValueTypeInteger, + .value.value.i = VA_SURFACE_ATTRIB_MEM_TYPE_DRM_PRIME_2, + }, + { + .type = VASurfaceAttribExternalBufferDescriptor, + .flags = VA_SURFACE_ATTRIB_SETTABLE, + .value.type = VAGenericValueTypePointer, + .value.value.p = &prime_desc, + } + }; + prime_desc.fourcc = va_fourcc; + prime_desc.width = src_fc->width; + prime_desc.height = src_fc->height; + prime_desc.num_objects = desc->nb_objects; + for (i = 0; i < desc->nb_objects; ++i) { + prime_desc.objects[i].fd = desc->objects[i].fd; + prime_desc.objects[i].size = desc->objects[i].size; + prime_desc.objects[i].drm_format_modifier = + desc->objects[i].format_modifier; } - } - buffer_desc.num_planes = k; - if (format_desc->chroma_planes_swapped && - buffer_desc.num_planes == 3) { - FFSWAP(uint32_t, buffer_desc.pitches[1], buffer_desc.pitches[2]); - FFSWAP(uint32_t, buffer_desc.offsets[1], buffer_desc.offsets[2]); + prime_desc.num_layers = desc->nb_layers; + for (i = 0; i < desc->nb_layers; ++i) { + prime_desc.layers[i].drm_format = desc->layers[i].format; + prime_desc.layers[i].num_planes = desc->layers[i].nb_planes; + for (j = 0; j < desc->layers[i].nb_planes; ++j) { + prime_desc.layers[i].object_index[j] = + desc->layers[i].planes[j].object_index; + prime_desc.layers[i].offset[j] = desc->layers[i].planes[j].offset; + prime_desc.layers[i].pitch[j] = desc->layers[i].planes[j].pitch; + } + + if (format_desc->chroma_planes_swapped && + desc->layers[i].nb_planes == 3) { + FFSWAP(uint32_t, prime_desc.layers[i].pitch[1], + prime_desc.layers[i].pitch[2]); + FFSWAP(uint32_t, prime_desc.layers[i].offset[1], + prime_desc.layers[i].offset[2]); + } + } + + /* + * We can query for PRIME_2 support with vaQuerySurfaceAttributes, but that + * that needs the config_id which we don't have here . Both Intel and + * Gallium seem to do the correct error checks, so lets just try the + * PRIME_2 import first. + */ + vas = vaCreateSurfaces(dst_dev->display, format_desc->rt_format, + src->width, src->height, &surface_id, 1, + prime_attrs, FF_ARRAY_ELEMS(prime_attrs)); + if (vas != VA_STATUS_SUCCESS) + src_vafc->prime_2_import_unsupported = 1; } - vas = vaCreateSurfaces(dst_dev->display, format_desc->rt_format, - src->width, src->height, - &surface_id, 1, - attrs, FF_ARRAY_ELEMS(attrs)); + if (!use_prime2 || vas != VA_STATUS_SUCCESS) { + int k; + unsigned long buffer_handle; + VASurfaceAttribExternalBuffers buffer_desc; + VASurfaceAttrib buffer_attrs[2] = { + { + .type = VASurfaceAttribMemoryType, + .flags = VA_SURFACE_ATTRIB_SETTABLE, + .value.type = VAGenericValueTypeInteger, + .value.value.i = VA_SURFACE_ATTRIB_MEM_TYPE_DRM_PRIME, + }, + { + .type = VASurfaceAttribExternalBufferDescriptor, + .flags = VA_SURFACE_ATTRIB_SETTABLE, + .value.type = VAGenericValueTypePointer, + .value.value.p = &buffer_desc, + } + }; + + buffer_handle = desc->objects[0].fd; + buffer_desc.pixel_format = va_fourcc; + buffer_desc.width = src_fc->width; + buffer_desc.height = src_fc->height; + buffer_desc.data_size = desc->objects[0].size; + buffer_desc.buffers = &buffer_handle; + buffer_desc.num_buffers = 1; + buffer_desc.flags = 0; + + k = 0; + for (i = 0; i < desc->nb_layers; i++) { + for (j = 0; j < desc->layers[i].nb_planes; j++) { + buffer_desc.pitches[k] = desc->layers[i].planes[j].pitch; + buffer_desc.offsets[k] = desc->layers[i].planes[j].offset; + ++k; + } + } + buffer_desc.num_planes = k; + + if (format_desc->chroma_planes_swapped && + buffer_desc.num_planes == 3) { + FFSWAP(uint32_t, buffer_desc.pitches[1], buffer_desc.pitches[2]); + FFSWAP(uint32_t, buffer_desc.offsets[1], buffer_desc.offsets[2]); + } + + vas = vaCreateSurfaces(dst_dev->display, format_desc->rt_format, + src->width, src->height, + &surface_id, 1, + buffer_attrs, FF_ARRAY_ELEMS(buffer_attrs)); + } if (vas != VA_STATUS_SUCCESS) { av_log(dst_fc, AV_LOG_ERROR, "Failed to create surface from DRM " "object: %d (%s).\n", vas, vaErrorStr(vas));