[syslog-ng] [Bug 278] Many memory leak problems happen when do configuration reloading

bugzilla at bugzilla.balabit.com bugzilla at bugzilla.balabit.com
Tue Jun 3 05:24:13 CEST 2014


https://bugzilla.balabit.com/show_bug.cgi?id=278





--- Comment #15 from Xufeng Zhang <xufeng.zhang at windriver.com>  2014-06-03 05:24:12 ---
(In reply to comment #14)
> For the above 3) case, I think there is a memory leak problem in
> modules/affile/logproto-file-writer.c:
> Take the following scenario, we call log_proto_file_writer_post()
> to post a message, if we have consumed the previous data and after register
> the new message, consumed variable will be TRUE, this indicates we will allocate
> a new self->line_buffer->str buffer in log_writer_flush() after we return from
> log_proto_file_writer_post(), but there are cases that the new message has not been
> copied to the partial buffer and freed in log_proto_file_writer_flush(), one case is
> log_proto_file_writer_flush() return LPS_ERROR when I/O error happened, I think this
> is the reason why my previous patch "Don't allocate a new buffer if fails to consume current item"
> partially works.
> How would we handle this situation?
> 

My above statement is not correct, from my test result, the below memory leak:
3).
==888== 77,312 bytes in 151 blocks are definitely lost in loss record 607 of 607
==888==    at 0x4A05F58: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==888==    by 0x3AE26478F4: g_malloc (in /lib64/libglib-2.0.so.0.2600.0)
==888==    by 0x3AE1E41838: log_writer_flush (logwriter.c:949)
==888==    by 0x3AE1E418EC: log_writer_work_perform (logwriter.c:129)
==888==    by 0x3AE1E41957: log_writer_io_flush_output (logwriter.c:209)
==888==    by 0x3AE1E5EC29: iv_run_tasks (iv_task.c:48)
==888==    by 0x3AE1E60E7B: iv_main (iv_main_posix.c:106)
==888==    by 0x3AE1E41E16: main_loop_run (mainloop.c:737)
==888==    by 0x4017A0: main (main.c:267)
logwriter.c:949
self->line_buffer->str = g_malloc(self->line_buffer->allocated_len);

is still come from the text-client, not from file-writer, so the fix
("logwriter: still free the unconsumed item during reloading configuration") on git
branch f/memory-leak-fixes only partly works.


-- 
Configure bugmail: https://bugzilla.balabit.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are watching all bug changes.


More information about the syslog-ng mailing list