Comment 65 for bug 1235649

Revision history for this message
Steve Langasek (vorlon) wrote : Re: [Bug 1235649] Re: uevent spam causes libdbus client code in session upstart to consume massive amounts of memory on Ubuntu Touch

On Tue, Oct 15, 2013 at 08:19:53AM -0000, James Hunt wrote:
> Once again, the biggest offender by a huge margin is dbus:

> ==15329== 1,153,679,208 bytes in 12,025 blocks are still reachable in loss record 208 of 208
> ==15329== at 0x402BB88: realloc (in /usr/lib/valgrind/vgpreload_memcheck-x86-linux.so)
> ==15329== by 0x408E778: dbus_realloc (dbus-memory.c:678)
> ==15329== by 0x408F08C: reallocate_for_length (dbus-string.c:349)
> ==15329== by 0x408F12B: set_length (dbus-string.c:390)
> ==15329== by 0x408F1A9: open_gap (dbus-string.c:411)
> ==15329== by 0x408FA67: copy (dbus-string.c:1198)
> ==15329== by 0x408FCCE: _dbus_string_copy_len (dbus-string.c:1368)
> ==15329== by 0x408DCF8: marshal_len_followed_by_bytes (dbus-marshal-basic.c:726)
> ==15329== by 0x408DD95: marshal_string (dbus-marshal-basic.c:758)
> ==15329== by 0x408DF60: _dbus_marshal_write_basic (dbus-marshal-basic.c:840)
> ==15329== by 0x4075CD7: _dbus_type_writer_write_basic_no_typecode (dbus-marshal-recursive.c:1601)
> ==15329== by 0x40768BB: _dbus_type_writer_write_basic (dbus-marshal-recursive.c:2323)

Yet this backtrace is still deep in the dbus internals and shows us nothing
about who the actual caller is that's causing the leak. Can you please run
this with --num-callers=20 so we can see why these strings are actually
being allocated?