[FFmpeg-devel] [PATCH 00/14] avcodec/vvcdec: support subpicture

Nuo Mi nuomi2021 at gmail.com
Tue Mar 19 16:22:50 EET 2024


On Tue, Mar 19, 2024 at 6:18 PM Frank Plowman <post at frankplowman.com> wrote:

> On 18/03/2024 14:16, Nuo Mi wrote:
> > see introductions here:
> https://dashif.org/docs/VVC%20HLS%20overview%20.pdf
> >
> > Frank Plowman (1):
> >   avcodec/vvcdec: support rectangular single-slice subpics
> >
> > Nuo Mi (13):
> >   avcodec/vvcdec: NoBackwardPredFlag, only check active pictures
> >   avcodec/cbs_h266: fix sh_collocated_from_l0_flag and
> >     sh_collocated_ref_idx infer
> >   avcodec/vvcdec: derive subpic postion for PPS
> >   avcodec/vvcdec: ff_vvc_decode_neighbour, support subpicture
> >   avcodec/vvcdec: misc, rename x_ctb, y_ctb, ctu_x, ctu_y to rx, ry to
> >     avoid misleading
> >   avcodec/vvcdec: refact out deblock_is_boundary
> >   avcodec/vvcdec: deblock, support subpicture
> >   avcodec/vvcdec: refact, movie the lc->sc assignment to task_run_stage
> >     to simplify the code
> >   avcodec/vvcdec: sao, refact out tile_edge arrays
> >   avcodec/vvcdec: sao, support subpicture
> >   avcodec/vvcdec: alf, support subpicture
> >   avcodec/vvcdec: mvs, support subpicture
> >   avcodec/vvcdec: inter prediction, support subpicture
> >
> >  libavcodec/cbs_h266_syntax_template.c |  37 ++--
> >  libavcodec/vvc/vvc_ctu.c              |  12 +-
> >  libavcodec/vvc/vvc_ctu.h              |   6 +-
> >  libavcodec/vvc/vvc_filter.c           | 233 +++++++++++++-------------
> >  libavcodec/vvc/vvc_filter.h           |   6 +-
> >  libavcodec/vvc/vvc_inter.c            |  79 ++++++---
> >  libavcodec/vvc/vvc_mvs.c              |  35 ++--
> >  libavcodec/vvc/vvc_ps.c               | 150 ++++++++++++++---
> >  libavcodec/vvc/vvc_ps.h               |   4 +
> >  libavcodec/vvc/vvc_thread.c           |  66 ++------
> >  10 files changed, 384 insertions(+), 244 deletions(-)
> >
> > --
> > 2.25.1
>
> Thank you.
>
> When compiled without a sanitizer, all subpicture conformance bitstreams
> decode correctly, however when compiled with UBSAN/ASAN this set causes
> a lot of errors for me.
>
👍, good catch.
Will check it.

>
> With UBSAN, I get errors similar to:
>
> libavcodec/vvc/vvc_ctu.c:1553:13: runtime error: member access within
> misaligned address 0xffffb52dbd6c for type 'av_alias64', which requires
> 8 byte alignment
> 0xffffb52dbd6c: note: pointer points here
>   f0 b3 9a 99 ff ff 00 00  00 be 2d b5 ff ff 00 00  c4 a6 fb d5 aa aa 00
> 00  c0 f8 2d b5 ff ff 00 00
>
> on the following sequences:
> * SUBPIC_B_HUAWEI
> * SUBPIC_C_ERICSSON
> * SUBPIC_E_MediaTek
> * LMCS_B_Dolby
> * CodingTools_E_Tencent
>
> With ASAN, I get errors similar to:
>
> =================================================================
> ==67289==ERROR: AddressSanitizer: heap-buffer-overflow on address
> 0x60300000067a at pc 0xaaaad00d21a0 bp 0xfffffba438f0 sp 0xfffffba43908
> READ of size 2 at 0x60300000067a thread T0
>     #0 0xaaaad00d219c in subpic_tiles
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:374:12
>     #1 0xaaaad00d1048 in pps_subpic_slice
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:406:5
>     #2 0xaaaad00cf2cc in pps_single_slice_per_subpic
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:419:13
>     #3 0xaaaad00ce8e0 in pps_rect_slice
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:464:9
>     #4 0xaaaad00cd4f8 in pps_slice_map
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:499:9
>     #5 0xaaaad00caf64 in pps_derive
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:543:11
>     #6 0xaaaad00cac00 in pps_alloc
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:576:11
>     #7 0xaaaad00c64f8 in decode_pps
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:598:11
>     #8 0xaaaad00c4444 in decode_ps
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:630:11
>     #9 0xaaaad00c3e70 in ff_vvc_decode_frame_ps
> /home/frank/FFmpeg/libavcodec/vvc/vvc_ps.c:866:11
> ...
>
> 0x60300000067a is located 0 bytes after 10-byte region
> [0x603000000670,0x60300000067a)
>
> on the following sequences:
> * SUBPIC_C_ERICSSON
> * SUBPIC_D_ERICSSON
>
> Note SUBPIC_D_ERICSSON has the ASAN error but no UBSAN error.
>
> Cheers,
> Frank
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>
> To unsubscribe, visit link above, or email
> ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
>


More information about the ffmpeg-devel mailing list