From ffae3ca84f646530fe9236b0966b663aebb06717 Mon Sep 17 00:00:00 2001
From: Paul B Mahol <onemda@gmail.com>
Date: Thu, 6 Jun 2024 17:45:43 +0800
Subject: [PATCH] Fix CVE-2023-49528

Conflict:NA
Reference:2d9ed64859c9887d0504cd71dbd5b2c15e14251a
---
 libavfilter/af_dialoguenhance.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)

diff --git a/libavfilter/af_dialoguenhance.c b/libavfilter/af_dialoguenhance.c
index 1762ea7..0924c3e 100644
--- a/libavfilter/af_dialoguenhance.c
+++ b/libavfilter/af_dialoguenhance.c
@@ -96,12 +96,13 @@ static int config_input(AVFilterLink *inlink)
     if (!s->window)
         return AVERROR(ENOMEM);
 
-    s->in_frame       = ff_get_audio_buffer(inlink, s->fft_size * 4);
-    s->center_frame   = ff_get_audio_buffer(inlink, s->fft_size * 4);
-    s->out_dist_frame = ff_get_audio_buffer(inlink, s->fft_size * 4);
-    s->windowed_frame = ff_get_audio_buffer(inlink, s->fft_size * 4);
-    s->windowed_out   = ff_get_audio_buffer(inlink, s->fft_size * 4);
-    s->windowed_prev  = ff_get_audio_buffer(inlink, s->fft_size * 4);
+    s->in_frame       = ff_get_audio_buffer(inlink, (s->fft_size + 2) * 2);
+    s->center_frame   = ff_get_audio_buffer(inlink, (s->fft_size + 2) * 2);
+    s->out_dist_frame = ff_get_audio_buffer(inlink, (s->fft_size + 2) * 2);
+    s->windowed_frame = ff_get_audio_buffer(inlink, (s->fft_size + 2) * 2);
+    s->windowed_out   = ff_get_audio_buffer(inlink, (s->fft_size + 2) * 2);
+    s->windowed_prev  = ff_get_audio_buffer(inlink, (s->fft_size + 2) * 2);
+
     if (!s->in_frame || !s->windowed_out || !s->windowed_prev ||
         !s->out_dist_frame || !s->windowed_frame || !s->center_frame)
         return AVERROR(ENOMEM);
@@ -250,6 +251,7 @@ static int de_stereo(AVFilterContext *ctx, AVFrame *out)
     float *right_osamples  = (float *)out->extended_data[1];
     float *center_osamples = (float *)out->extended_data[2];
     const int offset = s->fft_size - s->overlap;
+    const int nb_samples = FFMIN(s->overlap, s->in->nb_samples);
     float vad;
 
     // shift in/out buffers
@@ -258,8 +260,8 @@ static int de_stereo(AVFilterContext *ctx, AVFrame *out)
     memmove(left_out, &left_out[s->overlap], offset * sizeof(float));
     memmove(right_out, &right_out[s->overlap], offset * sizeof(float));
 
-    memcpy(&left_in[offset], left_samples, s->overlap * sizeof(float));
-    memcpy(&right_in[offset], right_samples, s->overlap * sizeof(float));
+    memcpy(&left_in[offset], left_samples, nb_samples * sizeof(float));
+    memcpy(&right_in[offset], right_samples, nb_samples * sizeof(float));
     memset(&left_out[offset], 0, s->overlap * sizeof(float));
     memset(&right_out[offset], 0, s->overlap * sizeof(float));
 
-- 
2.33.0