ryanzryu opened a new pull request, #27775:
URL: https://github.com/apache/doris/pull/27775
## Proposed changes
I got a core stack like this:
#0 _mm_loadu_si128(long long __vector(2) const*) (__P=0x576ff) at
/var/local/ldb-toolchain/lib/gcc/x86_64-linux-gnu/11/include/emmintrin.h:703
#1 inline_memcpy (size=77472, src_=<optimized out>, dst_=0x7f76a5252d00) at
/data/doris-1.x/be/src/glibc-compatibility/memcpy/memcpy_x86_64.cpp:187
#2 memcpy (dst=0x7f76a5252d00, src=<optimized out>, size=size@entry=77472)
at /data/doris-1.x/be/src/glibc-compatibility/memcpy/memcpy_x86_64.cpp:219
#3 0x0000557c566d3abd in memcpy (__len=77472, __src=<optimized out>,
__dest=<optimized out>) at
/var/local/ldb-toolchain/usr/include/x86_64-linux-gnu/bits/string_fortified.h:34
#4 doris::BufferedReader::_read_once (this=0x7f7647e98180,
position=<optimized out>, nbytes=<optimized out>, bytes_read=0x7f7a19088178,
out=<optimized out>)
at /data/doris-1.x/be/src/io/buffered_reader.cpp:137
#5 0x0000557c566d3c42 in doris::BufferedReader::readat
(this=0x7f7647e98180, position=358146, nbytes=77472, bytes_read=0x7f7a19088178,
out=0x7f76a5252d00)
at /data/doris-1.x/be/src/io/buffered_reader.cpp:94
#6 0x0000557c56817a27 in doris::ArrowFile::ReadAt (this=0x7f765b4ec4b0,
position=<optimized out>, nbytes=77472, out=0x7f76a5252d00)
at /data/doris-1.x/be/src/exec/arrow/arrow_reader.cpp:228
#7 0x0000557c5d07a6d0 in ?? ()
#8 0x0000557c5dd4e724 in orc::SeekableFileInputStream::Next(void const**,
int*) ()
#9 0x0000557c5dd4fd0c in orc::DecompressionStream::readBuffer(bool) ()
#10 0x0000557c5dd4fa4f in orc::DecompressionStream::readHeader() ()
#11 0x0000557c5dd4fe30 in orc::DecompressionStream::Next(void const**, int*)
()
#12 0x0000557c5dd5aed1 in
orc::StringDirectColumnReader::next(orc::ColumnVectorBatch&, unsigned long,
char*) ()
#13 0x0000557c5dd5a207 in
orc::StructColumnReader::next(orc::ColumnVectorBatch&, unsigned long, char*) ()
#14 0x0000557c5dd4679b in orc::RowReaderImpl::next(orc::ColumnVectorBatch&)
()
#15 0x0000557c5d07b243 in ?? ()
#16 0x0000557c5ce7403f in
arrow::RecordBatchReader::ReadAll(std::vector<std::shared_ptr<arrow::RecordBatch>,
std::allocator<std::shared_ptr<arrow::RecordBatch> > >*) ()
#17 0x0000557c57bed85d in doris::ORCReaderWrap::read_batches
(this=0x7f7a0f5a0400, batches=..., current_group=<optimized out>)
at /var/local/ldb-toolchain/include/c++/11/bits/shared_ptr_base.h:1290
#18 0x0000557c5681907f in doris::ArrowReaderWrap::prefetch_batch
(this=0x7f7a0f5a0400) at /data/doris-1.x/be/src/exec/arrow/arrow_reader.cpp:183
#19 0x0000557c5e7933e0 in execute_native_thread_routine ()
#20 0x00007f7aa0215ea5 in start_thread () from /lib64/libpthread.so.0
#21 0x00007f7aa05289fd in clone () from /lib64/libc.so.6
I gdb the core file and found be core because of BufferedReader::_read_once
call memcpy using null pointor _buffer.
I have read the relevant code, and I couldn't find any place that sets the
_buffer to a null pointer except for BufferedReader::close(). The close()
function is only called when BufferedReader is destructed. I didn't find any
place where BufferedReader continues to be used after it is destructed.
Therefore, I will temporarily add a null pointer protection before memcpy.
## Further comments
If this is a relatively large or complex change, kick off the discussion at
[[email protected]](mailto:[email protected]) by explaining why you
chose the solution you did and what alternatives you considered, etc...
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
