From: Zhao Zhili <quinkblack@foxmail.com> To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org> Subject: Re: [FFmpeg-devel] [PATCH v5] avcodec/libx264: fix extradata when config annexb=0 Date: Sun, 17 Mar 2024 11:46:47 +0800 Message-ID: <tencent_EF003968FE7CD2ACA2387DC0F9B6006BA60A@qq.com> (raw) In-Reply-To: <AS8P250MB0744483C1C44F78A9F1A20168F292@AS8P250MB0744.EURP250.PROD.OUTLOOK.COM> > On Mar 14, 2024, at 21:14, Andreas Rheinhardt <andreas.rheinhardt@outlook.com> wrote: > > Zhao Zhili: >> From: Zhao Zhili <zhilizhao@tencent.com> >> >> --- >> v5: Decode chroma_format_idc directly instead of ff_h264_decode_seq_parameter_set >> v4: Fix missing SEI in set_avcc_extradata >> v3: Remove unnecessary inclusion >> >> libavcodec/libx264.c | 162 ++++++++++++++++++++++++++++++++++++------- >> 1 file changed, 138 insertions(+), 24 deletions(-) >> >> diff --git a/libavcodec/libx264.c b/libavcodec/libx264.c >> index 10d646bd76..64cefb5fb0 100644 >> --- a/libavcodec/libx264.c >> +++ b/libavcodec/libx264.c >> @@ -38,6 +38,7 @@ >> #include "packet_internal.h" >> #include "atsc_a53.h" >> #include "sei.h" >> +#include "golomb.h" >> >> #include <x264.h> >> #include <float.h> >> @@ -865,6 +866,140 @@ static int convert_pix_fmt(enum AVPixelFormat pix_fmt) >> return 0; >> } >> >> +static int save_sei(AVCodecContext *avctx, x264_nal_t *nal) >> +{ >> + X264Context *x4 = avctx->priv_data; >> + >> + av_log(avctx, AV_LOG_INFO, "%s\n", nal->p_payload + 25); >> + x4->sei_size = nal->i_payload; >> + x4->sei = av_malloc(x4->sei_size); >> + if (!x4->sei) >> + return AVERROR(ENOMEM); >> + >> + memcpy(x4->sei, nal->p_payload, nal->i_payload); >> + >> + return 0; >> +} >> + >> +static int set_avcc_extradata(AVCodecContext *avctx, x264_nal_t *nal, int nnal) >> +{ >> + X264Context *x4 = avctx->priv_data; >> + x264_nal_t *sps_nal = NULL; >> + x264_nal_t *pps_nal = NULL; >> + uint8_t *p, *sps; >> + int ret; >> + >> + /* We know it's in the order of SPS/PPS/SEI, but it's not documented in x264 API. >> + * The x264 param i_sps_id implies there is a single pair of SPS/PPS. >> + */ >> + for (int i = 0; i < nnal; i++) { >> + switch (nal[i].i_type) { >> + case NAL_SPS: >> + sps_nal = &nal[i]; >> + break; >> + case NAL_PPS: >> + pps_nal = &nal[i]; >> + break; >> + case NAL_SEI: >> + ret = save_sei(avctx, &nal[i]); >> + if (ret < 0) >> + return ret; >> + break; >> + } >> + } >> + if (!sps_nal || !pps_nal) >> + return AVERROR_EXTERNAL; >> + >> + avctx->extradata_size = sps_nal->i_payload + pps_nal->i_payload + 7; >> + avctx->extradata = av_mallocz(avctx->extradata_size + AV_INPUT_BUFFER_PADDING_SIZE); >> + if (!avctx->extradata) >> + return AVERROR(ENOMEM); >> + >> + // Now create AVCDecoderConfigurationRecord >> + p = avctx->extradata; >> + // Skip size part >> + sps = sps_nal->p_payload + 4; >> + *p++ = 1; // version >> + *p++ = sps[1]; // AVCProfileIndication >> + *p++ = sps[2]; // profile_compatibility >> + *p++ = sps[3]; // AVCLevelIndication >> + *p++ = 0xFF; >> + *p++ = 0xE0 | 0x01; // 3 bits reserved (111) + 5 bits number of sps >> + memcpy(p, sps_nal->p_payload + 2, sps_nal->i_payload - 2); >> + // Make sps has AV_INPUT_BUFFER_PADDING_SIZE padding, so it can be used >> + // with GetBitContext >> + sps = p + 2; >> + p += sps_nal->i_payload - 2; >> + *p++ = 1; >> + memcpy(p, pps_nal->p_payload + 2, pps_nal->i_payload - 2); >> + p += pps_nal->i_payload - 2; >> + >> + if (sps[3] != 66 && sps[3] != 77 && sps[3] != 88) { >> + GetBitContext gbc; >> + int chroma_format_idc; >> + int bit_depth_luma_minus8, bit_depth_chroma_minus8; >> + >> + /* It's not possible to have emulation prevention byte before >> + * bit_depth_chroma_minus8 due to the range of sps id, chroma_format_idc >> + * and so on. So we can read directly without need to escape emulation >> + * prevention byte. > > There is a slight problem here: You ignore the possibility of > profile_idc being zero. After all, all values not being used are merely > reserved and can be used at any time (although I doubt that they will > ever use this). Shall we ignore this? > The above check (and presumably the ISOBMFF spec for this) uses an > opt-out check for profile, whereas the H.264 spec actually uses an > opt-in check. I just wanted to mention it, feel free to ignore it. > >> + * >> + * +4 to skip until sps id. >> + */ >> + init_get_bits8(&gbc, sps + 4, sps_nal->i_payload - 4 - 4); >> + // Skip sps id >> + get_ue_golomb_31(&gbc); >> + chroma_format_idc = get_ue_golomb_31(&gbc); >> + if (chroma_format_idc == 3) >> + skip_bits1(&gbc); >> + bit_depth_luma_minus8 = get_ue_golomb_31(&gbc); >> + bit_depth_chroma_minus8 = get_ue_golomb_31(&gbc); >> + >> + *p++ = 0xFC | chroma_format_idc; >> + *p++ = 0xF8 | bit_depth_luma_minus8; >> + *p++ = 0xF8 | bit_depth_chroma_minus8; >> + *p++ = 0; >> + } >> + av_assert0(avctx->extradata + avctx->extradata_size >= p); > > Does this have to be an av_assert0? Switched to assert2 in v6. > >> + avctx->extradata_size = p - avctx->extradata; >> + >> + return 0; >> +} >> + >> +static int set_extradata(AVCodecContext *avctx) >> +{ >> + X264Context *x4 = avctx->priv_data; >> + x264_nal_t *nal; >> + uint8_t *p; >> + int nnal, s; >> + >> + s = x264_encoder_headers(x4->enc, &nal, &nnal); >> + if (s < 0) >> + return AVERROR_EXTERNAL; >> + >> + if (!x4->params.b_annexb) >> + return set_avcc_extradata(avctx, nal, nnal); >> + >> + avctx->extradata = p = av_mallocz(s + AV_INPUT_BUFFER_PADDING_SIZE); >> + if (!p) >> + return AVERROR(ENOMEM); >> + >> + for (int i = 0; i < nnal; i++) { >> + /* Don't put the SEI in extradata. */ >> + if (nal[i].i_type == NAL_SEI) { >> + s = save_sei(avctx, &nal[i]); >> + if (s < 0) >> + return s; >> + continue; >> + } >> + memcpy(p, nal[i].p_payload, nal[i].i_payload); >> + p += nal[i].i_payload; >> + } >> + avctx->extradata_size = p - avctx->extradata; >> + >> + return 0; >> +} >> + >> #define PARSE_X264_OPT(name, var)\ >> if (x4->var && x264_param_parse(&x4->params, name, x4->var) < 0) {\ >> av_log(avctx, AV_LOG_ERROR, "Error parsing option '%s' with value '%s'.\n", name, x4->var);\ >> @@ -1233,30 +1368,9 @@ FF_ENABLE_DEPRECATION_WARNINGS >> return AVERROR_EXTERNAL; >> >> if (avctx->flags & AV_CODEC_FLAG_GLOBAL_HEADER) { >> - x264_nal_t *nal; >> - uint8_t *p; >> - int nnal, s, i; >> - >> - s = x264_encoder_headers(x4->enc, &nal, &nnal); >> - avctx->extradata = p = av_mallocz(s + AV_INPUT_BUFFER_PADDING_SIZE); >> - if (!p) >> - return AVERROR(ENOMEM); >> - >> - for (i = 0; i < nnal; i++) { >> - /* Don't put the SEI in extradata. */ >> - if (nal[i].i_type == NAL_SEI) { >> - av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25); >> - x4->sei_size = nal[i].i_payload; >> - x4->sei = av_malloc(x4->sei_size); >> - if (!x4->sei) >> - return AVERROR(ENOMEM); >> - memcpy(x4->sei, nal[i].p_payload, nal[i].i_payload); >> - continue; >> - } >> - memcpy(p, nal[i].p_payload, nal[i].i_payload); >> - p += nal[i].i_payload; >> - } >> - avctx->extradata_size = p - avctx->extradata; >> + ret = set_extradata(avctx); >> + if (ret < 0) >> + return ret; >> } >> >> cpb_props = ff_encode_add_cpb_side_data(avctx); > > You need a golomb configure dependency for get_ue_golomb_31(). Select golomb in configure and use conditional compile in case of only x262 is enabled. I think atsc_a53 has the same problem but it’s unrelated to current patch. > > - Andreas > > _______________________________________________ > ffmpeg-devel mailing list > ffmpeg-devel@ffmpeg.org > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel > > To unsubscribe, visit link above, or email > ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe". _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org https://ffmpeg.org/mailman/listinfo/ffmpeg-devel To unsubscribe, visit link above, or email ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
prev parent reply other threads:[~2024-03-17 3:47 UTC|newest] Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top 2024-03-13 13:36 Zhao Zhili 2024-03-14 13:14 ` Andreas Rheinhardt 2024-03-17 3:46 ` Zhao Zhili [this message]
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=tencent_EF003968FE7CD2ACA2387DC0F9B6006BA60A@qq.com \ --to=quinkblack@foxmail.com \ --cc=ffmpeg-devel@ffmpeg.org \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel This inbox may be cloned and mirrored by anyone: git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git # If you have public-inbox 1.1+ installed, you may # initialize and index your mirror using the following commands: public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \ ffmpegdev@gitmailbox.com public-inbox-index ffmpegdev Example config snippet for mirrors. AGPL code for this site: git clone https://public-inbox.org/public-inbox.git