Opened 14 years ago

Closed 13 years ago

Last modified 13 years ago

#394 closed defect (fixed)

SIGSEGV crash in h264 decoder

Reported by: dario@… Owned by: lorenm@…
Priority: normal Component: libavcodec
Version: unspecified Severity: normal
Keywords: Cc: diego@…
Blocked By: Blocking:
Reproduced by developer: Analyzed by developer:


I’ve got a SIGSEGV decoding h264, the stream was generated by x264.

The following stream was recorded live and begins from the middle:

the following command line reproduces the fault (the source code is updated
from cvs as of 09/05/2005 08:13pm GMT-3).

$ ./ffmpeg -i dump.SIGSEGV.h264 -f null dump.SIGSEGV.h264.null

ffmpeg version CVS, build 3211266, Copyright (c) 2000-2004 Fabrice Bellard

configuration: --disable-ffserver --disable-ffplay --enable-memalign-hack --

enable-mingw32 --source-path=c:/cygwin/home/dario/ffmpeg

built on Sep 5 2005 19:18:54, gcc: 3.2.3 (mingw special 20030504-1)

Input #0, h264, from 'dump.SIGSEGV.h264':

Duration: N/A, bitrate: N/A

Stream #0.0: Video: h264, yuv420p, 320x240, 10.00 fps

Output #0, null, to 'dump.SIGSEGV.h264.null':

Stream #0.0: Video: rawvideo, yuv420p, 320x240, 10.00 fps, q=2-31, 200 kb/s

Stream mapping:

Stream #0.0 -> #0.0

frame= 104 q=0.0 size= 0kB time=10.4 bitrate= 0.0kbits/s

Here are the requested bug reporting data:

(gdb) r -i dump.SIGSEGV.h264 -f null dump.SIGSEGV.h264.null

Starting program: C:\cygwin\home\dario\ffmpeg/ffmpeg_g.exe -i
dump.SIGSEGV.h264 -f null dump.SIGSEGV.h264.null

Program received signal SIGSEGV, Segmentation fault.

decode_frame (avctx=0x3de270, data=0x22f9d0, data_size=0x22f898,

buf=0xe66ac8 "", buf_size=895) at h264.c:7553

7553 if(h->delayed_pic[i]->key_frame



(gdb) bt

#0 decode_frame (avctx=0x3de270, data=0x22f9d0, data_size=0x22f898,

buf=0xe66ac8 "", buf_size=895) at h264.c:7553

#1 0x0045d69a in avcodec_decode_video (avctx=0x3de270, picture=0x22f9d0,

got_picture_ptr=0x22f898, buf=0xe66ac8 "", buf_size=895) at utils.c:625

#2 0x0040cda9 in output_packet (ist=0xd5ced0, ist_index=0,

ost_table=0xd31d60, nb_ostreams=1, pkt=0x22fb30) at ffmpeg.c:1266

#3 0x00406026 in av_encode (output_files=0x7100f0, nb_output_files=1,

input_files=0x710000, nb_input_files=1, stream_maps=0x710140,

nb_stream_maps=0) at ffmpeg.c:2102

#4 0x004048e5 in main (argc=6, argv=0x3d2de0) at ffmpeg.c:4520

(gdb) disass $pc-32 $pc+32

Dump of assembler code from 0x5d53d6 to 0x5d5416:

0x5d53d6 <decode_frame+438>: out %al,(%dx)

0x5d53d7 <decode_frame+439>: add %eax,(%eax)

0x5d53d9 <decode_frame+441>: inc %esi

0x5d53da <decode_frame+442>: mov %esi,0xffffffd8(%ebp)

0x5d53dd <decode_frame+445>: test %ecx,%ecx

0x5d53df <decode_frame+447>: jne 0x5d53e8 <decode_frame+456>

0x5d53e1 <decode_frame+449>: movl $0x1,0x50(%eax)

0x5d53e8 <decode_frame+456>: mov 0x1ee18(%ebx),%eax

0x5d53ee <decode_frame+462>: xor %ecx,%ecx

0x5d53f0 <decode_frame+464>: test %eax,%eax

0x5d53f2 <decode_frame+466>: je 0x5d541a <decode_frame+506>

0x5d53f4 <decode_frame+468>: mov %eax,%esi

0x5d53f6 <decode_frame+470>: mov 0x30(%esi),%edx

0x5d53f9 <decode_frame+473>: test %edx,%edx

0x5d53fb <decode_frame+475>: jne 0x5d5407 <decode_frame+487>

0x5d53fd <decode_frame+477>: mov 0xe4(%esi),%edi

0x5d5403 <decode_frame+483>: test %edi,%edi

0x5d5405 <decode_frame+485>: jne 0x5d540e <decode_frame+494>

0x5d5407 <decode_frame+487>: movl $0x1,0xffffffd4(%ebp)

0x5d540e <decode_frame+494>: inc %ecx

0x5d540f <decode_frame+495>: mov 0x1ee18(%ebx,%ecx,4),%esi

End of assembler dump.

(gdb) info all-registers

eax 0xda51b0 14307760

ecx 0x11 17

edx 0x0 0

ebx 0xd84fc0 14176192

esp 0x22f7a0 0x22f7a0

ebp 0x22f7d8 0x22f7d8

esi 0x1 1

edi 0x32 50

eip 0x5d53f6 0x5d53f6

eflags 0x10202 66050

cs 0x1b 27

ss 0x23 35

ds 0x23 35

es 0x23 35

fs 0x3b 59

gs 0x0 0

st0 -nan(0x6e6e6e6e6e6e6e6e) (raw 0xffff6e6e6e6e6e6e6e6e)

st1 -nan(0x6e6e6e6e6e6e6e6e) (raw 0xffff6e6e6e6e6e6e6e6e)

st2 -nan(0x6e68566b686b6c72) (raw 0xffff6e68566b686b6c72)

st3 -nan(0x6e00680056006b) (raw 0xffff006e00680056006b)

st4 0 (raw 0xffff0000000000000000)

st5 0 (raw 0xffff0000000000000000)

st6 0 (raw 0xffff0000000000000000)

st7 0 (raw 0xffff0000000000000000)

fctrl 0xffff037f -64641

fstat 0xffff0020 -65504

ftag 0xffffffff -1

fiseg 0x1b 27

fioff 0x4065e9 4220393

foseg 0xffff0023 -65501

fooff 0xd5cf08 14012168

fop 0x1ca 458

(gdb) l h264.c:7550

7545 }


7547 while(h->delayed_pic[pics]) pics++;

7548 h->delayed_pic[pics++] = cur;

7549 if(cur->reference == 0)

7550 cur->reference = 1;


7552 for(i=0; h->delayed_pic[i]; i++)

7553 if(h->delayed_pic[i]->key_frame

poc==0) ß-------- ERROR in here!!


7554 cross_idr = 1;


7556 out = h->delayed_pic[0];

7557 for(i=1; h->delayed_pic[i] && !h->delayed_pic[i]->key_frame;

7558 if(h->delayed_pic[i]->poc < out->poc){

7559 out = h->delayed_pic[i];

7560 out_idx = i;

7561 }


7563 out_of_order = !cross_idr && prev && out->poc < prev->poc;

7564 if(prev && pics <= s->avctx->has_b_frames)

(gdb) info locals

out = (Picture *) 0xffffffff

cur = (Picture *) 0xda61a0

prev = (Picture *) 0xda6338

out_idx = 0

pics = 16

cross_idr = 0

dropped_frame = 0

s = (MpegEncContext? *) 0xd84fc0

buf_index = 895

buf_index = 895

(gdb) out h

(H264Context *) 0xda51b0

(gdb) out h->delayed_pic

{0xbaadf00d <repeats 16 times>}

Attachments (1)

dump.SIGSEGV.h264 (668.0 KB) - added by dario@… 14 years ago.
Live partial h264 stream dump

Download all attachments as: .zip

Change History (7)

Changed 14 years ago by dario@…

Live partial h264 stream dump

comment:1 Changed 14 years ago by dario@…

command line:
./ffmpeg -i dump.SIGSEGV.h264 -f null .

comment:2 Changed 14 years ago by diego@…

  • Cc diego@… added
  • op_sys changed from MinGW to All
  • Owner changed from diego@… to lorenm@…

Loren, can you look into this? I have verified this crash to still occur on
Linux x86 and PPC with latest CVS.

comment:3 Changed 14 years ago by lorenm@…

  • Status changed from new to assigned

No longer crashes.
This particular stream dump is broken beyond repair: about 25% of all frames are
missing, including most of the I-frames.
The frame reordering code is still wrong in the presence of any dropped frames
at all, so I'll look into heuristics for salvaging less thoroughly broken streams.

comment:4 Changed 13 years ago by dominik@…

I recommend closing as FIXED.

comment:5 Changed 13 years ago by dario@…

Thanks a lot guys. I believe the provided patch have fixed the problem, there
were no more crashes since then.
Thanks again.

comment:6 Changed 13 years ago by diego@…

  • Resolution set to fixed
  • Status changed from assigned to closed

This issue is now FIXED.

Note: See TracTickets for help on using tickets.