[FFmpeg-devel] [PATCH] avformat/rtpdec: int overflow in start_time_realtime

[Martin Storsjö]

On Wed, 8 Jan 2025, jon at jonb.org wrote:

> From: Jonathan Baudanza <jon at jonb.org>
>
> This was previously adjusted by me in 6b3f9c2e92b.
> Unfortunately, I traded one integer overflow bug for
> another.
>
> Currently, NTP timestamps that exceed INT64_MAX
> (~Jan 20, 1968) will cause an overflow when passed
> to av_rescale.
>
> This patch replaces av_rescale, which operates on
> int64_t, with ff_parse_ntp_time, which operates on
> uint64_t. This will give the correct values for
> timestamps back around the NTP epoch and present day
> timestamps.
> ---
> libavformat/rtsp.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/libavformat/rtsp.c b/libavformat/rtsp.c
> index c48fa26d90..5ea471b40c 100644
> --- a/libavformat/rtsp.c
> +++ b/libavformat/rtsp.c
> @@ -2320,7 +2320,7 @@ redo:
>                 }
>                 // Make real NTP start time available in AVFormatContext
>                 if (s->start_time_realtime == AV_NOPTS_VALUE) {
> -                    s->start_time_realtime = av_rescale (rtpctx->first_rtcp_ntp_time, 1000000, 1LL << 32) - NTP_OFFSET_US;
> +                    s->start_time_realtime = ff_parse_ntp_time(rtpctx->first_rtcp_ntp_time) - NTP_OFFSET_US;
>                     if (rtpctx->st) {
>                         s->start_time_realtime -=
>                             av_rescale_q (rtpctx->rtcp_ts_offset, rtpctx->st->time_base, AV_TIME_BASE_Q);
> -- 
> 2.41.0

Looks ok - I've managed to reproduce this now.

I'll apply this, and backport it to the 7.1 branch.

// Martin