1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2025-08-04 22:03:09 +02:00

avcodec/mpegvideo: Allocate dc_val for each encoder slice

This fixes data races (which are UB) in the MPEG-4 and H.263+
encoder when predicting DC values; these encoders unconditionally
read values from the line above the current line and only check
lateron (via first_slice_line) whether said prediction can be used
at all. It will also allow to remove said checks (by setting the
entries to 1024 upon opening a new slice).

The vsynth{1,2,3,_lena}-mpeg4-thread FATE tests were affected by this:
https://fate.ffmpeg.org/report.cgi?slot=x86_64-archlinux-gcc-tsan-slices&time=20250613002615

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This commit is contained in:
Andreas Rheinhardt
2025-06-13 10:21:13 +02:00
parent eb597d9a1e
commit c84a03d752
2 changed files with 13 additions and 3 deletions

View File

@ -179,6 +179,7 @@ static void backup_duplicate_context(MpegEncContext *bak, MpegEncContext *src)
COPY(block); COPY(block);
COPY(start_mb_y); COPY(start_mb_y);
COPY(end_mb_y); COPY(end_mb_y);
COPY(dc_val);
COPY(ac_val); COPY(ac_val);
#undef COPY #undef COPY
} }
@ -235,7 +236,7 @@ av_cold int ff_mpv_init_context_frame(MpegEncContext *s)
s->avctx->active_thread_type & FF_THREAD_SLICE) ? s->avctx->active_thread_type & FF_THREAD_SLICE) ?
s->avctx->thread_count : 1; s->avctx->thread_count : 1;
BufferPoolContext *const pools = &s->buffer_pools; BufferPoolContext *const pools = &s->buffer_pools;
int y_size, c_size, yc_size, i, mb_array_size, mv_table_size, x, y; int y_size, c_size, yc_size, mb_array_size, mv_table_size, x, y;
int mb_height; int mb_height;
if (s->encoding && s->avctx->slices) if (s->encoding && s->avctx->slices)
@ -337,6 +338,9 @@ av_cold int ff_mpv_init_context_frame(MpegEncContext *s)
} }
if (s->h263_pred || s->h263_aic || !s->encoding) { if (s->h263_pred || s->h263_aic || !s->encoding) {
// When encoding, each slice (and therefore each thread)
// gets its own ac_val and dc_val buffers in order to avoid
// races.
size_t allslice_yc_size = yc_size * (s->encoding ? nb_slices : 1); size_t allslice_yc_size = yc_size * (s->encoding ? nb_slices : 1);
if (s->out_format == FMT_H263) { if (s->out_format == FMT_H263) {
/* ac values */ /* ac values */
@ -349,10 +353,15 @@ av_cold int ff_mpv_init_context_frame(MpegEncContext *s)
// MN: we need these for error resilience of intra-frames // MN: we need these for error resilience of intra-frames
// Allocating them unconditionally for decoders also means // Allocating them unconditionally for decoders also means
// that we don't need to reinitialize when e.g. h263_aic changes. // that we don't need to reinitialize when e.g. h263_aic changes.
if (!FF_ALLOC_TYPED_ARRAY(s->dc_val_base, yc_size))
// y_size and therefore yc_size is always odd; allocate one element
// more for each encoder slice in order to be able to align each slice's
// dc_val to four in order to use aligned stores when cleaning dc_val.
allslice_yc_size += s->encoding * nb_slices;
if (!FF_ALLOC_TYPED_ARRAY(s->dc_val_base, allslice_yc_size))
return AVERROR(ENOMEM); return AVERROR(ENOMEM);
s->dc_val = s->dc_val_base + s->b8_stride + 1; s->dc_val = s->dc_val_base + s->b8_stride + 1;
for (i = 0; i < yc_size; i++) for (size_t i = 0; i < allslice_yc_size; ++i)
s->dc_val_base[i] = 1024; s->dc_val_base[i] = 1024;
} }

View File

@ -545,6 +545,7 @@ static av_cold int init_slice_buffers(MPVMainEncContext *const m)
} }
if (s2->c.ac_val) { if (s2->c.ac_val) {
s2->c.dc_val += offset + i;
s2->c.ac_val += offset; s2->c.ac_val += offset;
offset += yc_size; offset += yc_size;
} }