From patchwork Fri Dec 31 16:00:42 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-Patchwork-Submitter: Romain Beauxis X-Patchwork-Id: 32967 Delivered-To: ffmpegpatchwork2@gmail.com Received: by 2002:a6b:cd86:0:0:0:0:0 with SMTP id d128csp14071534iog; Fri, 31 Dec 2021 08:00:56 -0800 (PST) X-Google-Smtp-Source: ABdhPJz9MBVuqPWCdIhAG06yD7HmC3kNWIMqZd8lqbNmQPx9g4OGkt6KdiJPYiqQEzSvEWPxMYMl X-Received: by 2002:a17:906:56c9:: with SMTP id an9mr14768670ejc.755.1640966456331; Fri, 31 Dec 2021 08:00:56 -0800 (PST) ARC-Seal: i=1; a=rsa-sha256; t=1640966456; cv=none; d=google.com; s=arc-20160816; b=lFFxLpI6tYyAeCsARh4jZmuX0eEieRijiF7mV3w0u6kkiPGeGXnBtQJ/tddTsu0z3j PYUItNFmnxSoICiYvR7uZ3wz+WcwFFTIUxHzZk66aA2WGVLZVcVnVlNS+jk1ljZIfYXX q1tQ1mobBj6J7gOQ1ICH01f6GnaIXcexmJNU2Y6cAuitkfHmRbc2A5LHGSL50D8u/VE3 JkcjegJElRAA4OdG8K8tlRYDoWdn8yaAjgLkdxccq+zJHg/e/stuyTfUHhtOfWZkqEka rL2y+kRs1LB2/Xiji//kichUYRveSKHB4oGb4hmuCqF37FWYpBKNlzClcLIRIb2jfVxC v1Tg== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:content-transfer-encoding:cc:reply-to :list-subscribe:list-help:list-post:list-archive:list-unsubscribe :list-id:precedence:subject:to:date:message-id:mime-version:from :delivered-to; bh=6h014sy7qWHyf2byNPyOoHkkHQ78rUVB3/MfWV2c6Bc=; b=R54xaaUVQJ7FABwvUEPasXayV4Jue3dmfhEdTg4EcoouSUeRA90xskfQEHeTHWKgTE l2tdnWSEkOI9oe0V6LbpC6jKk6mYC+u/TA7eFAN7RJIvG2UPOTIbpV0HHnFownNNqus2 KqXZ/rv3sbmFINdhdhM9kl/+1JxTjS2DHVmqBrp2PJEuv6yFNoya1AYvmjWyUM6zHRP1 YbCbzOLn/fLEr6gJiMDc6d+xmo6vMe8rhr4uFXLcNU0PH6r2dvwAIX3PAUy5ev2VMje/ 8G1ZurlG9FwbM902VMjVKet5pT1QGK2PijvnbSPi5uqnf1qCzynhTbQgNeTwbjhSQTb5 jaSw== 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 Return-Path: Received: from ffbox0-bg.mplayerhq.hu (ffbox0-bg.ffmpeg.org. [79.124.17.100]) by mx.google.com with ESMTP id w18si14976426edl.552.2021.12.31.08.00.56; Fri, 31 Dec 2021 08:00:56 -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 Received: from [127.0.1.1] (localhost [127.0.0.1]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTP id 23F9368B040; Fri, 31 Dec 2021 18:00:53 +0200 (EET) X-Original-To: ffmpeg-devel@ffmpeg.org Delivered-To: ffmpeg-devel@ffmpeg.org Received: from mail-qt1-f171.google.com (mail-qt1-f171.google.com [209.85.160.171]) by ffbox0-bg.mplayerhq.hu (Postfix) with ESMTPS id 5BBF46805AE for ; Fri, 31 Dec 2021 18:00:46 +0200 (EET) Received: by mail-qt1-f171.google.com with SMTP id v4so21887569qtk.0 for ; Fri, 31 Dec 2021 08:00:46 -0800 (PST) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:from:content-transfer-encoding:mime-version :subject:message-id:date:cc:to; bh=W4A4UJMwwG7lw6PJ+qxap6ZLdQ2zvkl369DeShDqJMI=; b=Po+xsz31OVGCuMZPKSSrnDKPQGU0Ns0taMXBJdzEclXtwrVsXn76KpvGzrDVRBQMkn wBPhWx7IG4CbgM/zmf0bTfmMN/NRRfK/y7NJ4nAwJHjbAI/lJL34ZV55ZuLtRrT49zBc ufTQULBh1w7gE3z0VuwMeeVEnc1B9qVfpHYNNXnQGgUwiv92zYcqxqGycs5CTviuvzso uyLPeWysKCgdlzd7VJa+21mYARTTtzzEK74g+e6o/Rv0jnWhXQoyYIk7HW29pSogEAX9 ktZTWtHZJlsK2QcQkvCPwXVFdIH75vaAYUL6IUgt1H0rouQ5b/cbzquK9mC8HvAAt0GF jfdg== X-Gm-Message-State: AOAM532vAt9dkQN4rsg9bLXdKwc2O3xlSzvfNkzaMDDkBPsh8wljubC1 OpJnhdZc/IPDu848uL7xD9nBbJOEYZE0zw== X-Received: by 2002:ac8:59c9:: with SMTP id f9mr30778276qtf.204.1640966444417; Fri, 31 Dec 2021 08:00:44 -0800 (PST) Received: from smtpclient.apple ([172.58.171.222]) by smtp.gmail.com with ESMTPSA id d4sm22132118qkn.79.2021.12.31.08.00.43 (version=TLS1_2 cipher=ECDHE-ECDSA-AES128-GCM-SHA256 bits=128/128); Fri, 31 Dec 2021 08:00:44 -0800 (PST) From: Romain Beauxis Mime-Version: 1.0 (Mac OS X Mail 15.0 \(3693.40.0.1.81\)) Message-Id: <01FD4E04-7D9C-4284-9AB4-FF8AA3E6C20F@rastageeks.org> Date: Fri, 31 Dec 2021 10:00:42 -0600 To: ffmpeg-devel@ffmpeg.org X-Mailer: Apple Mail (2.3693.40.0.1.81) Subject: [FFmpeg-devel] [PATCH v7 2/3] libavdevice/avfoundation.m: Replace mutex-based concurrency handling in avfoundation.m by a thread-safe fifo queue with maximum length 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 Cc: Thilo Borgmann , Aman Karmani Errors-To: ffmpeg-devel-bounces@ffmpeg.org Sender: "ffmpeg-devel" X-TUID: s40XK1Rdwt2L * Use a CMSimpleQueueEnqueue with maximum length to queue and process incoming audio and video frames. * Log avfoundation errors. * Use AVERROR_EXTERNAL instead of AVERROR(EIO) in avfoundation errors. Signed-off-by: Romain Beauxis — This is the second patch of a series of 3 that fix, cleanup and enhance the avfoundation implementation for libavdevice. These patches come from an actual user-facing application relying on libavdevice’s implementation of avfoundation audio input. Without them, Avfoundation is practically unusable as it will: * Refuse to process certain specific audio input format that are actually returned by the OS for some users (packed PCM audio) * Drop audio frames, resulting in corrupted audio input. This might have been unnoticed with video frames but this makes avfoundation essentially unusable for audio. The patches are now being included in our production build so they are tested and usable in production. This patch fixes the concurrency model. Avfoundation runs its own producing thread to send produced frames and ffmpeg runs its own thread to consume them. The existing implementation stores the last transmitted frame and uses a mutex to avoid concurrent access. However, this leads to situations where upcoming frames can be dropped if the ffmpeg thread is acessing the latest frame. This happens even when the thread would otherwise catch up and process frames fast enought. This patches changes this implementation to use a buffer queue with a max queue length and encapsulated thread-safety. This greatly simplifies the logic of the calling code and gives the consuming thread a chance to process all frames concurrently to the producing thread while avoiding memory leaks. Changes: v2: None v3: Switched queue implementation to CMSimpleQueue v4: None v5: Fix indentation/wrapping v6: Fix audio/video frame queue cleanup logic v7: Enhance avfoundation error reporting: add human-readable description, use AVERROR_EXTERNAL instead of AVERROR(EIO) libavdevice/avfoundation.m | 196 ++++++++++++++++++------------------- 1 file changed, 96 insertions(+), 100 deletions(-) diff --git a/libavdevice/avfoundation.m b/libavdevice/avfoundation.m index 738cd93375..c740745fce 100644 --- a/libavdevice/avfoundation.m +++ b/libavdevice/avfoundation.m @@ -26,7 +26,7 @@ */ #import -#include +#import #include "libavutil/channel_layout.h" #include "libavutil/pixdesc.h" @@ -39,6 +39,11 @@ #include "libavutil/imgutils.h" #include "avdevice.h" +#define av_log_avfoundation_error(str, err) \ + av_log(s, AV_LOG_ERROR, "Avfoundation: %s, %s\n", str, \ + [[[NSError errorWithDomain:NSOSStatusErrorDomain code:err userInfo:nil] localizedDescription] UTF8String] \ + ) + static const int avf_time_base = 1000000; static const AVRational avf_time_base_q = { @@ -80,13 +85,12 @@ { AV_PIX_FMT_NONE, 0 } }; +#define MAX_QUEUED_FRAMES 10 + typedef struct { AVClass* class; - int frames_captured; - int audio_frames_captured; - pthread_mutex_t frame_lock; id avf_delegate; id avf_audio_delegate; @@ -122,8 +126,8 @@ AVCaptureSession *capture_session; AVCaptureVideoDataOutput *video_output; AVCaptureAudioDataOutput *audio_output; - CMSampleBufferRef current_frame; - CMSampleBufferRef current_audio_frame; + CMSimpleQueueRef audio_frames_queue; + CMSimpleQueueRef video_frames_queue; AVCaptureDevice *observed_device; #if !TARGET_OS_IPHONE && __MAC_OS_X_VERSION_MIN_REQUIRED >= 1070 @@ -132,16 +136,6 @@ int observed_quit; } AVFContext; -static void lock_frames(AVFContext* ctx) -{ - pthread_mutex_lock(&ctx->frame_lock); -} - -static void unlock_frames(AVFContext* ctx) -{ - pthread_mutex_unlock(&ctx->frame_lock); -} - /** FrameReciever class - delegate for AVCaptureSession */ @interface AVFFrameReceiver : NSObject @@ -219,17 +213,8 @@ - (void) captureOutput:(AVCaptureOutput *)captureOutput didOutputSampleBuffer:(CMSampleBufferRef)videoFrame fromConnection:(AVCaptureConnection *)connection { - lock_frames(_context); - - if (_context->current_frame != nil) { - CFRelease(_context->current_frame); - } - - _context->current_frame = (CMSampleBufferRef)CFRetain(videoFrame); - - unlock_frames(_context); - - ++_context->frames_captured; + CFRetain(videoFrame); + CMSimpleQueueEnqueue(_context->video_frames_queue, videoFrame); } @end @@ -263,17 +248,8 @@ - (void) captureOutput:(AVCaptureOutput *)captureOutput didOutputSampleBuffer:(CMSampleBufferRef)audioFrame fromConnection:(AVCaptureConnection *)connection { - lock_frames(_context); - - if (_context->current_audio_frame != nil) { - CFRelease(_context->current_audio_frame); - } - - _context->current_audio_frame = (CMSampleBufferRef)CFRetain(audioFrame); - - unlock_frames(_context); - - ++_context->audio_frames_captured; + CFRetain(audioFrame); + CMSimpleQueueEnqueue(_context->audio_frames_queue, audioFrame); } @end @@ -288,6 +264,30 @@ static void destroy_context(AVFContext* ctx) [ctx->avf_delegate release]; [ctx->avf_audio_delegate release]; + CMSampleBufferRef frame; + + if (ctx->video_frames_queue) { + frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->video_frames_queue); + while (frame) { + CFRelease(frame); + frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->video_frames_queue); + } + + CFRelease(ctx->video_frames_queue); + ctx->video_frames_queue = NULL; + } + + if (ctx->audio_frames_queue) { + frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->audio_frames_queue); + while (frame) { + CFRelease(frame); + frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->audio_frames_queue); + } + + CFRelease(ctx->audio_frames_queue); + ctx->audio_frames_queue = NULL; + } + ctx->capture_session = NULL; ctx->video_output = NULL; ctx->audio_output = NULL; @@ -298,12 +298,6 @@ static void destroy_context(AVFContext* ctx) AudioConverterDispose(ctx->audio_converter); ctx->audio_converter = NULL; } - - pthread_mutex_destroy(&ctx->frame_lock); - - if (ctx->current_frame) { - CFRelease(ctx->current_frame); - } } static void parse_device_name(AVFormatContext *s) @@ -631,18 +625,18 @@ static int get_video_config(AVFormatContext *s) } // Take stream info from the first frame. - while (ctx->frames_captured < 1) { + while (CMSimpleQueueGetCount(ctx->video_frames_queue) < 1) { CFRunLoopRunInMode(kCFRunLoopDefaultMode, 0.1, YES); } - lock_frames(ctx); + CMSampleBufferRef frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->video_frames_queue); ctx->video_stream_index = stream->index; avpriv_set_pts_info(stream, 64, 1, avf_time_base); - image_buffer = CMSampleBufferGetImageBuffer(ctx->current_frame); - block_buffer = CMSampleBufferGetDataBuffer(ctx->current_frame); + image_buffer = CMSampleBufferGetImageBuffer(frame); + block_buffer = CMSampleBufferGetDataBuffer(frame); if (image_buffer) { image_buffer_size = CVImageBufferGetEncodedSize(image_buffer); @@ -658,10 +652,7 @@ static int get_video_config(AVFormatContext *s) stream->codecpar->format = ctx->pixel_format; } - CFRelease(ctx->current_frame); - ctx->current_frame = nil; - - unlock_frames(ctx); + CFRelease(frame); return 0; } @@ -681,27 +672,27 @@ static int get_audio_config(AVFormatContext *s) } // Take stream info from the first frame. - while (ctx->audio_frames_captured < 1) { + while (CMSimpleQueueGetCount(ctx->audio_frames_queue) < 1) { CFRunLoopRunInMode(kCFRunLoopDefaultMode, 0.1, YES); } - lock_frames(ctx); + CMSampleBufferRef frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->audio_frames_queue); ctx->audio_stream_index = stream->index; avpriv_set_pts_info(stream, 64, 1, avf_time_base); - format_desc = CMSampleBufferGetFormatDescription(ctx->current_audio_frame); + format_desc = CMSampleBufferGetFormatDescription(frame); const AudioStreamBasicDescription *input_format = CMAudioFormatDescriptionGetStreamBasicDescription(format_desc); if (!input_format) { - unlock_frames(ctx); + CFRelease(frame); av_log(s, AV_LOG_ERROR, "audio format not available\n"); return 1; } if (input_format->mFormatID != kAudioFormatLinearPCM) { - unlock_frames(ctx); + CFRelease(frame); av_log(s, AV_LOG_ERROR, "only PCM audio format are supported at the moment\n"); return 1; } @@ -781,16 +772,13 @@ static int get_audio_config(AVFormatContext *s) if (must_convert) { OSStatus ret = AudioConverterNew(input_format, &output_format, &ctx->audio_converter); if (ret != noErr) { - unlock_frames(ctx); - av_log(s, AV_LOG_ERROR, "Error while allocating audio converter\n"); + CFRelease(frame); + av_log_avfoundation_error("error while creating audio converter", ret); return 1; } } - CFRelease(ctx->current_audio_frame); - ctx->current_audio_frame = nil; - - unlock_frames(ctx); + CFRelease(frame); return 0; } @@ -808,8 +796,6 @@ static int avf_read_header(AVFormatContext *s) ctx->num_video_devices = [devices count] + [devices_muxed count]; - pthread_mutex_init(&ctx->frame_lock, NULL); - #if !TARGET_OS_IPHONE && __MAC_OS_X_VERSION_MIN_REQUIRED >= 1070 CGGetActiveDisplayList(0, NULL, &num_screens); #endif @@ -1010,6 +996,21 @@ static int avf_read_header(AVFormatContext *s) // Initialize capture session ctx->capture_session = [[AVCaptureSession alloc] init]; + OSStatus ret; + ret = CMSimpleQueueCreate(kCFAllocatorDefault, MAX_QUEUED_FRAMES, &ctx->video_frames_queue); + + if (ret != noErr) { + av_log_avfoundation_error("error while creating frame queue", ret); + goto fail; + } + + ret = CMSimpleQueueCreate(kCFAllocatorDefault, MAX_QUEUED_FRAMES, &ctx->audio_frames_queue); + + if (ret != noErr) { + av_log_avfoundation_error("error while creating frame queue", ret); + goto fail; + } + if (video_device && add_video_device(s, video_device)) { goto fail; } @@ -1039,7 +1040,8 @@ static int avf_read_header(AVFormatContext *s) fail: [pool release]; destroy_context(ctx); - return AVERROR(EIO); + av_log(s, AV_LOG_ERROR, "Error while opening AVfoundation capture session\n"); + return AVERROR_EXTERNAL; } static int copy_cvpixelbuffer(AVFormatContext *s, @@ -1088,38 +1090,35 @@ static int copy_cvpixelbuffer(AVFormatContext *s, static int avf_read_packet(AVFormatContext *s, AVPacket *pkt) { OSStatus ret; + int status; AVFContext* ctx = (AVFContext*)s->priv_data; do { - CVImageBufferRef image_buffer; - CMBlockBufferRef block_buffer; - lock_frames(ctx); - - if (ctx->current_frame != nil) { - int status; + if (1 <= CMSimpleQueueGetCount(ctx->video_frames_queue)) { int length = 0; - - image_buffer = CMSampleBufferGetImageBuffer(ctx->current_frame); - block_buffer = CMSampleBufferGetDataBuffer(ctx->current_frame); + CMSampleBufferRef video_frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->video_frames_queue); + CVImageBufferRef image_buffer = CMSampleBufferGetImageBuffer(video_frame);; + CMBlockBufferRef block_buffer = CMSampleBufferGetDataBuffer(video_frame); if (image_buffer != nil) { length = (int)CVPixelBufferGetDataSize(image_buffer); } else if (block_buffer != nil) { length = (int)CMBlockBufferGetDataLength(block_buffer); } else { - unlock_frames(ctx); + CFRelease(video_frame); return AVERROR(EINVAL); } - if (av_new_packet(pkt, length) < 0) { - unlock_frames(ctx); - return AVERROR(EIO); + status = av_new_packet(pkt, length); + if (status < 0) { + CFRelease(video_frame); + return status; } CMItemCount count; CMSampleTimingInfo timing_info; - if (CMSampleBufferGetOutputSampleTimingInfoArray(ctx->current_frame, 1, &timing_info, &count) == noErr) { + if (CMSampleBufferGetOutputSampleTimingInfoArray(video_frame, 1, &timing_info, &count) == noErr) { AVRational timebase_q = av_make_q(1, timing_info.presentationTimeStamp.timescale); pkt->pts = pkt->dts = av_rescale_q(timing_info.presentationTimeStamp.value, timebase_q, avf_time_base_q); } @@ -1133,18 +1132,18 @@ static int avf_read_packet(AVFormatContext *s, AVPacket *pkt) status = 0; ret = CMBlockBufferCopyDataBytes(block_buffer, 0, pkt->size, pkt->data); if (ret != kCMBlockBufferNoErr) { - status = AVERROR(EIO); + av_log_avfoundation_error("error while copying buffer data", ret); + status = AVERROR_EXTERNAL; } } - CFRelease(ctx->current_frame); - ctx->current_frame = nil; + CFRelease(video_frame); if (status < 0) { - unlock_frames(ctx); return status; } - } else if (ctx->current_audio_frame != nil) { - CMBlockBufferRef block_buffer = CMSampleBufferGetDataBuffer(ctx->current_audio_frame); + } else if (1 <= CMSimpleQueueGetCount(ctx->audio_frames_queue)) { + CMSampleBufferRef audio_frame = (CMSampleBufferRef)CMSimpleQueueDequeue(ctx->audio_frames_queue); + CMBlockBufferRef block_buffer = CMSampleBufferGetDataBuffer(audio_frame); size_t input_size = CMBlockBufferGetDataLength(block_buffer); int buffer_size = input_size / ctx->audio_buffers; @@ -1170,8 +1169,9 @@ static int avf_read_packet(AVFormatContext *s, AVPacket *pkt) if (ret != kCMBlockBufferNoErr) { av_free(input_buffer); - unlock_frames(ctx); - return AVERROR(EIO); + CFRelease(audio_frame); + av_log_avfoundation_error("error while accessing audio buffer data", ret); + return AVERROR_EXTERNAL; } } @@ -1188,23 +1188,25 @@ static int avf_read_packet(AVFormatContext *s, AVPacket *pkt) av_free(input_buffer); if (ret != noErr) { - unlock_frames(ctx); - return AVERROR(EIO); + CFRelease(audio_frame); + av_log_avfoundation_error("error while converting audio data", ret); + return AVERROR_EXTERNAL; } pkt->size = output_buffer.mBuffers[0].mDataByteSize; } else { ret = CMBlockBufferCopyDataBytes(block_buffer, 0, pkt->size, pkt->data); if (ret != kCMBlockBufferNoErr) { - unlock_frames(ctx); - return AVERROR(EIO); + CFRelease(audio_frame); + av_log_avfoundation_error("error while copying audio data", ret); + return AVERROR_EXTERNAL; } } CMItemCount count; CMSampleTimingInfo timing_info; - if (CMSampleBufferGetOutputSampleTimingInfoArray(ctx->current_audio_frame, 1, &timing_info, &count) == noErr) { + if (CMSampleBufferGetOutputSampleTimingInfoArray(audio_frame, 1, &timing_info, &count) == noErr) { AVRational timebase_q = av_make_q(1, timing_info.presentationTimeStamp.timescale); pkt->pts = pkt->dts = av_rescale_q(timing_info.presentationTimeStamp.value, timebase_q, avf_time_base_q); } @@ -1212,21 +1214,15 @@ static int avf_read_packet(AVFormatContext *s, AVPacket *pkt) pkt->stream_index = ctx->audio_stream_index; pkt->flags |= AV_PKT_FLAG_KEY; - CFRelease(ctx->current_audio_frame); - ctx->current_audio_frame = nil; - - unlock_frames(ctx); + CFRelease(audio_frame); } else { pkt->data = NULL; - unlock_frames(ctx); if (ctx->observed_quit) { return AVERROR_EOF; } else { return AVERROR(EAGAIN); } } - - unlock_frames(ctx); } while (!pkt->data); return 0;