libwebm: handle unknown file length

Change-Id: I6ad22ce302e40a4ec882b4f9ac59d5ca7f6489d7
This commit is contained in:
matthewjheaney 2011-02-01 16:29:02 -08:00
parent a977a2b536
commit 7a8b013f16

View File

@ -35,8 +35,7 @@ long long mkvparser::ReadUInt(IMkvReader* pReader, long long pos, long& len)
long long total, available;
status = pReader->Length(&total, &available);
assert(status >= 0);
assert(total > 0);
assert(available <= total);
assert((total < 0) || (available <= total));
assert(pos < available);
assert((available - pos) >= 1); //assume here max u-int len is 8
#endif
@ -101,8 +100,7 @@ long long mkvparser::GetUIntLength(
int status = pReader->Length(&total, &available);
assert(status >= 0);
assert(total >= 0);
assert(available <= total);
assert((total < 0) || (available <= total));
if (pos >= available)
return pos; //too few bytes available
@ -229,12 +227,14 @@ float mkvparser::Unserialize4Float(
assert(pos >= 0);
#ifdef _DEBUG
long long total, available;
{
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
assert((pos + 4) <= available);
const long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
assert((pos + 4) <= available);
}
#endif
#if 0
@ -345,12 +345,14 @@ signed char mkvparser::Unserialize1SInt(
assert(pos >= 0);
#ifdef _DEBUG
long long total, available;
{
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr == 0);
assert(available <= total);
assert(pos < available);
const long status = pReader->Length(&total, &available);
assert(status == 0);
assert((total < 0) || (available <= total));
assert(pos < available);
}
#endif
signed char result;
@ -370,12 +372,14 @@ short mkvparser::Unserialize2SInt(
assert(pos >= 0);
#ifdef _DEBUG
long long total, available;
{
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
assert((pos + 2) <= available);
const long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
assert((pos + 2) <= available);
}
#endif
#if 0
@ -428,9 +432,9 @@ bool mkvparser::Match(
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
const long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
long len;
@ -473,9 +477,9 @@ bool mkvparser::Match(
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
long len;
@ -506,14 +510,13 @@ bool mkvparser::Match(
{
char c;
hr = pReader->Read(pos + i, 1, (unsigned char*)&c);
assert(hr == 0L);
status = pReader->Read(pos + i, 1, (unsigned char*)&c);
assert(status == 0); //TODO
val[i] = c;
if (c == '\0')
break;
}
val[size] = '\0';
@ -534,9 +537,9 @@ bool mkvparser::Match(
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
long len;
const long long id = ReadUInt(pReader, pos, len);
@ -564,8 +567,8 @@ bool mkvparser::Match(
buf = new (std::nothrow) unsigned char[buflen_];
assert(buf); //TODO
hr = pReader->Read(pos, buflen_, buf);
assert(hr == 0L);
status = pReader->Read(pos, buflen_, buf);
assert(status == 0); //TODO
buflen = buflen_;
@ -585,10 +588,12 @@ bool mkvparser::Match(
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
const long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
long idlen;
const long long id = ReadUInt(pReader, pos, idlen);
assert(id >= 0); //TODO
@ -635,9 +640,9 @@ bool mkvparser::Match(
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
const long status = pReader->Length(&total, &available);
assert(status >= 0);
assert((total < 0) || (available <= total));
long len;
const long long id = ReadUInt(pReader, pos, len);
@ -708,13 +713,13 @@ long long EBMLHeader::Parse(
long long total, available;
long hr = pReader->Length(&total, &available);
long status = pReader->Length(&total, &available);
if (hr < 0)
return hr;
if (status < 0) //error
return status;
pos = 0;
long long end = (1024 < available)? 1024: available;
long long end = (available >= 1024) ? 1024 : available;
for (;;)
{
@ -722,10 +727,10 @@ long long EBMLHeader::Parse(
while (pos < end)
{
hr = pReader->Read(pos, 1, &b);
status = pReader->Read(pos, 1, &b);
if (hr < 0)
return hr;
if (status < 0) //error
return status;
if (b == 0x1A)
break;
@ -735,15 +740,16 @@ long long EBMLHeader::Parse(
if (b != 0x1A)
{
if ((pos >= 1024) ||
(available >= total) ||
((total - available) < 5))
return -1;
if (pos >= 1024)
return E_FILE_FORMAT_INVALID; //don't bother looking anymore
if ((total >= 0) && ((total - available) < 5))
return E_FILE_FORMAT_INVALID;
return available + 5; //5 = 4-byte ID + 1st byte of size
}
if ((total - pos) < 5)
if ((total >= 0) && ((total - pos) < 5))
return E_FILE_FORMAT_INVALID;
if ((available - pos) < 5)
@ -756,16 +762,19 @@ long long EBMLHeader::Parse(
if (result < 0) //error
return result;
if (result == 0x0A45DFA3) //ReadId masks-off length indicator bits
if (result == 0x0A45DFA3) //EBML Header ID
{
assert(len == 4);
pos += len;
pos += len; //consume ID
break;
}
++pos; //throw away just the 0x1A byte, and try again
}
//pos designates start of size field
//get length of size field
long len;
long long result = GetUIntLength(pReader, pos, len);
@ -778,20 +787,24 @@ long long EBMLHeader::Parse(
assert(len > 0);
assert(len <= 8);
if ((total - pos) < len)
if ((total >= 0) && ((total - pos) < len))
return E_FILE_FORMAT_INVALID;
if ((available - pos) < len)
return pos + len; //try again later
//get the EBML header size
result = ReadUInt(pReader, pos, len);
if (result < 0) //error
return result;
pos += len; //consume u-int
pos += len; //consume size field
if ((total - pos) < result)
//pos now designates start of payload
if ((total >= 0) && ((total - pos) < result))
return E_FILE_FORMAT_INVALID;
if ((available - pos) < result)
@ -898,9 +911,19 @@ long long Segment::CreateInstance(
long long total, available;
long hr = pReader->Length(&total, &available);
assert(hr >= 0);
assert(available <= total);
const long status = pReader->Length(&total, &available);
if (status < 0) //error
return status;
if (available < 0)
return -1;
if ((total >= 0) && (available > total))
return -1;
const long long end = (total >= 0) ? total : available;
//TODO: this might need to be liberalized
//I would assume that in practice this loop would execute
//exactly once, but we allow for other elements (e.g. Void)
@ -909,8 +932,17 @@ long long Segment::CreateInstance(
//but in the splitter case over a network we should probably
//just give up early. We could for example decide only to
//execute this loop a maximum of, say, 10 times.
//TODO:
//There is an implied "give up early" by only parsing up
//to the available limit. We do do that, but only if the
//total file size is unknown. We could decide to always
//use what's available as our limit (irrespective of whether
//we happen to know the total file length). This would have
//as its sense "parse this much of the file before giving up",
//which a slightly different sense from "try to parse up to
//10 EMBL elements before giving up".
while (pos < total)
while (pos < end)
{
//Read ID
@ -920,14 +952,12 @@ long long Segment::CreateInstance(
if (result) //error, or too few available bytes
return result;
if ((pos + len) > total)
if ((pos + len) > end)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
return pos + len;
//TODO: if we liberalize the behavior of ReadUInt, we can
//probably eliminate having to use GetUIntLength here.
const long long id = ReadUInt(pReader, pos, len);
if (id < 0) //error
@ -942,31 +972,33 @@ long long Segment::CreateInstance(
if (result) //error, or too few available bytes
return result;
if ((pos + len) > total)
if ((pos + len) > end)
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
return pos + len;
//TODO: if we liberalize the behavior of ReadUInt, we can
//probably eliminate having to use GetUIntLength here.
long long size = ReadUInt(pReader, pos, len);
if (size < 0)
if (size < 0) //error
return size;
pos += len; //consume length of size of element
//Pos now points to start of payload
//Handle "unknown size" for live streaming of webm files.
const long long unknown_size = (1LL << (7 * len)) - 1;
if (id == 0x08538067) //Segment ID
{
//Handle "unknown size" for live streaming of webm files.
const long long unknown_size = (1LL << (7 * len)) - 1;
if (size == unknown_size)
size = total - pos;
else if ((pos + size) > total)
size = -1;
else if (total < 0)
size = -1;
else if ((pos + size) > end)
return E_FILE_FORMAT_INVALID;
pSegment = new (std::nothrow) Segment(pReader, pos, size);
@ -977,13 +1009,17 @@ long long Segment::CreateInstance(
return 0; //success
}
if ((pos + size) > total)
if (size == unknown_size)
return E_FILE_FORMAT_INVALID;
if ((pos + size) > end)
return E_FILE_FORMAT_INVALID;
pos += size; //consume payload
}
return E_FILE_FORMAT_INVALID; //there is no segment
//TODO: this might need to be liberalized. See comments above.
}
@ -996,15 +1032,20 @@ long long Segment::ParseHeaders()
const int status = m_pReader->Length(&total, &available);
assert(status == 0);
assert(total >= 0);
assert(available <= total);
assert((total < 0) || (available <= total));
const long long stop = m_start + m_size;
assert(stop <= total);
assert(m_pos <= stop);
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
assert((segment_stop < 0) || (total < 0) || (segment_stop <= total));
assert((segment_stop < 0) || (m_pos <= segment_stop));
while (m_pos < stop)
for (;;)
{
if ((total >= 0) && (m_pos >= total))
break;
if ((segment_stop >= 0) && (m_pos >= segment_stop))
break;
long long pos = m_pos;
const long long element_start = pos;
@ -1020,7 +1061,7 @@ long long Segment::ParseHeaders()
if (result > 0) //underflow (weird)
return (pos + 1);
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
@ -1049,7 +1090,7 @@ long long Segment::ParseHeaders()
if (result > 0) //underflow (weird)
return (pos + 1);
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > available)
@ -1066,7 +1107,7 @@ long long Segment::ParseHeaders()
//Pos now points to start of payload
if ((pos + size) > stop)
if ((segment_stop >= 0) && ((pos + size) > segment_stop))
return E_FILE_FORMAT_INVALID;
//We read EBML elements either in total or nothing at all.
@ -1130,7 +1171,7 @@ long long Segment::ParseHeaders()
m_pos = pos + size; //consume payload
}
assert(m_pos <= stop);
assert((segment_stop < 0) || (m_pos <= segment_stop));
if (m_pInfo == NULL) //TODO: liberalize this behavior
return E_FILE_FORMAT_INVALID;
@ -1393,13 +1434,18 @@ long Segment::LoadCluster(
if (status < 0) //error
return status;
assert(total >= 0);
assert(avail <= total);
assert((total < 0) || (avail <= total));
const long long stop = m_start + m_size;
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
while (m_pos < stop)
for (;;)
{
if ((total >= 0) && (m_pos >= total))
return 1; //no more clusters
if ((segment_stop >= 0) && (m_pos >= segment_stop))
return 1; //no more clusters
pos = m_pos;
//Read ID
@ -1421,7 +1467,7 @@ long Segment::LoadCluster(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -1454,7 +1500,7 @@ long Segment::LoadCluster(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -1477,7 +1523,7 @@ long Segment::LoadCluster(
//Pos now points to start of payload
if ((pos + size) > stop)
if ((segment_stop >= 0) && ((pos + size) > segment_stop))
return E_FILE_FORMAT_INVALID;
len = static_cast<long>(size);
@ -1531,7 +1577,7 @@ long Segment::LoadCluster(
--m_clusterPreloadCount;
m_pos = pos + size; //consume payload
assert(m_pos <= stop);
assert((segment_stop < 0) || (m_pos <= segment_stop));
status = pCluster->LoadBlockEntries(pos, len);
assert(status == 0); //TODO
@ -1541,7 +1587,7 @@ long Segment::LoadCluster(
}
m_pos = pos + size; //consume payload
assert(m_pos <= stop);
assert((segment_stop < 0) || (m_pos <= segment_stop));
if (Cluster::HasBlockEntries(this, idoff))
{
@ -1563,9 +1609,6 @@ long Segment::LoadCluster(
return 0; //we have a new cluster
}
}
assert(m_pos <= stop);
return 1; //no error, but no new cluster either
}
@ -1721,23 +1764,14 @@ long Segment::Load()
assert(m_clusters == NULL);
assert(m_clusterSize == 0);
assert(m_clusterCount == 0);
assert(m_size >= 0); //TODO: we now allow size < 0
//Outermost (level 0) segment object has been constructed,
//and pos designates start of payload. We need to find the
//inner (level 1) elements.
const long long stop = m_start + m_size;
#ifdef _DEBUG //TODO: this is really Microsoft-specific
{
long long total, available;
long hr = m_pReader->Length(&total, &available);
assert(hr >= 0);
assert(available >= total);
assert(stop <= total);
}
#endif
while (m_pos < stop)
{
long long pos = m_pos;
@ -2068,13 +2102,12 @@ long Segment::ParseCues(
if (status < 0) //error
return status;
assert(total >= 0);
assert(avail <= total);
assert((total < 0) || (avail <= total));
pos = m_start + off;
const long long element_start = pos;
const long long stop = m_start + m_size; //end of segment
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
if ((pos + 1) > avail)
{
@ -2093,7 +2126,7 @@ long Segment::ParseCues(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop) //not a very useful test
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -2107,7 +2140,7 @@ long Segment::ParseCues(
return E_FILE_FORMAT_INVALID;
pos += len; //consume ID
assert(pos < stop);
assert((segment_stop < 0) || (pos <= segment_stop));
//Read Size
@ -2128,7 +2161,7 @@ long Segment::ParseCues(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop) //not a very useful test
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -2143,12 +2176,13 @@ long Segment::ParseCues(
return 1; //done
pos += len; //consume length of size of element
assert((segment_stop < 0) || (pos <= segment_stop));
//Pos now points to start of payload
const long long element_stop = pos + size;
if (element_stop > stop)
if ((segment_stop >= 0) && (element_stop > segment_stop))
return E_FILE_FORMAT_INVALID;
len = static_cast<long>(size);
@ -2166,14 +2200,6 @@ long Segment::ParseCues(
element_size);
assert(m_pCues); //TODO
#if 0 //TODO
//we should do this incrementally as well
while (m_pCues->LoadCuePoint())
;
#endif
return 0; //success
}
#endif
@ -3348,14 +3374,16 @@ long Segment::ParseNext(
if (status < 0) //error
return status;
assert(total >= 0);
assert(avail <= total);
assert((total < 0) || (avail <= total));
const long long off_curr_ = pCurr->m_pos;
const long long off_curr = off_curr_ * ((off_curr_ < 0) ? -1 : 1);
pos = m_start + off_curr;
const long long stop = m_start + m_size; //end of segment
const long long segment_stop = (m_size < 0) ? -1 : m_start + m_size;
//interrogate curr cluster
{
if ((pos + 1) > avail)
@ -3375,7 +3403,7 @@ long Segment::ParseNext(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -3407,7 +3435,7 @@ long Segment::ParseNext(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -3422,13 +3450,13 @@ long Segment::ParseNext(
pos += len; //consume length of size of element
if ((pos + size) > stop)
if ((segment_stop >= 0) && ((pos + size) > segment_stop))
return E_FILE_FORMAT_INVALID;
//Pos now points to start of payload
pos += size; //consume payload (that is, the current cluster)
assert(pos <= stop);
assert((segment_stop < 0) || (pos <= segment_stop));
//By consuming the payload, we are assuming that the curr
//cluster isn't interesting. That is, we don't bother checking
@ -3440,12 +3468,22 @@ long Segment::ParseNext(
//pos now points to just beyond the last fully-loaded cluster
//Parse next cluster. This is strictly a parsing activity.
//Creation of a new cluster object happens later, after the
//parsing is done.
long long off_next = 0;
long long element_start = -1;
long long element_size = -1;
while (pos < stop)
for (;;)
{
if ((total >= 0) && (pos >= total))
break;
if ((segment_stop >= 0) && (pos >= segment_stop))
break;
if ((pos + 1) > avail)
{
len = 1;
@ -3463,7 +3501,7 @@ long Segment::ParseNext(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -3501,7 +3539,7 @@ long Segment::ParseNext(
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > stop)
if ((segment_stop >= 0) && ((pos + len) > segment_stop))
return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
@ -3522,7 +3560,7 @@ long Segment::ParseNext(
element_start = idpos;
const long long element_stop = pos + size;
if (element_stop > stop)
if ((segment_stop >= 0) && (element_stop > segment_stop))
return E_FILE_FORMAT_INVALID;
element_size = element_stop - element_start;
@ -3540,7 +3578,7 @@ long Segment::ParseNext(
}
pos += size; //consume payload
assert(pos <= stop);
assert((segment_stop < 0) || (pos <= segment_stop));
continue;
}
@ -3548,7 +3586,7 @@ long Segment::ParseNext(
if (id != 0x0F43B675) //Cluster ID
{
pos += size; //consume payload
assert(pos <= stop);
assert((segment_stop < 0) || (pos <= segment_stop));
continue;
}
@ -3565,16 +3603,18 @@ long Segment::ParseNext(
}
pos += size; //consume payload
assert(pos <= stop);
assert((segment_stop < 0) || (pos <= segment_stop));
}
if (off_next <= 0) //no next cluster found
return 1;
//We have parsed the next cluster, and can even guarantee
//that its payload is all available (IMkvReader::Length).
//All we need to do now is determine whether it has already
//be preloaded.
//that its payload is all available (via IMkvReader::Length).
//We have not created a cluster object yet. What we need
//to do now is determine whether it has already be preloaded
//(in which case, an object for this cluster has already been
//created), and if not, create a new cluster object.
Cluster** const ii = m_clusters + m_clusterCount;
Cluster** i = ii;
@ -5078,10 +5118,10 @@ long Cluster::LoadBlockEntries(long long& pos, long& len) const
if (status < 0) //error
return status;
assert(total >= 0);
assert(avail <= total);
assert((total < 0) || (avail <= total));
const long long segment_stop = m_pSegment->m_start + m_pSegment->m_size;
//TODO: restore this check, but account for m_size < 0:
//const long long segment_stop = m_pSegment->m_start + m_pSegment->m_size;
if (m_pos > 0) //at least partially loaded
{
@ -5125,8 +5165,8 @@ long Cluster::LoadBlockEntries(long long& pos, long& len) const
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > segment_stop)
return E_FILE_FORMAT_INVALID;
//if ((pos + len) > segment_stop)
// return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
@ -5160,8 +5200,8 @@ long Cluster::LoadBlockEntries(long long& pos, long& len) const
return E_BUFFER_NOT_FULL;
}
if ((pos + len) > segment_stop)
return E_FILE_FORMAT_INVALID;
//if ((pos + len) > segment_stop)
// return E_FILE_FORMAT_INVALID;
if ((pos + len) > avail)
return E_BUFFER_NOT_FULL;
@ -5273,7 +5313,7 @@ Cluster* Cluster::Parse(
{
assert(pSegment);
assert(off >= 0);
assert(off < pSegment->m_size);
//assert(off < pSegment->m_size);
//if (!HasBlockEntries(pSegment, off))
// return NULL;