[Bug libmudflap/51858] New: Some malloc'd inside libc is not being tracked

["pinskia at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51858

             Bug #: 51858
           Summary: Some malloc'd inside libc is not being tracked
    Classification: Unclassified
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libmudflap
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: pinskia@gcc.gnu.org


On mips64-linux-gnu with our compiled glibc, we get many failures in libmudflap
testsuite.
The backtrace for those are failures is:
#0  __lll_lock_wait_private (futex=0x2ad84960) at
../nptl/sysdeps/unix/sysv/linux/lowlevellock.c:33
#1  0x2ad41380 in _Unwind_Find_registered_FDE (pc=0x2adf41b7, bases=0x7ffdf1e8)
at ../sysdeps/generic/unwind-dw2-fde.c:1011
#2  0x2ad413bc in _Unwind_Find_FDE (pc=0x2adf41b7, bases=0x7ffdf1e8) at
../sysdeps/generic/unwind-dw2-fde-glibc.c:257
#3  0x2adf2b24 in uw_frame_state_for (context=context@entry=0x7ffdeee8,
fs=fs@entry=0x7ffdd680) at
/home/apinski/src/gcc-fsf/local/gcc/libgcc/unwind-dw2.c:1179
#4  0x2adf30b0 in uw_init_context_1 (context=context@entry=0x7ffdeee8,
outer_cfa=outer_cfa@entry=0x7ffdf330, outer_ra=0x2ad1377c,
outer_ra@entry=0x2adf41b8)
    at /home/apinski/src/gcc-fsf/local/gcc/libgcc/unwind-dw2.c:1500
#5  0x2adf41b8 in _Unwind_Backtrace (trace=0x2ad135c0 <backtrace_helper>,
trace_argument=0x7ffdf330) at
/home/apinski/src/gcc-fsf/local/gcc/libgcc/unwind.inc:283
#6  0x2ad1377c in __GI___backtrace (array=<optimized out>, size=6) at
../sysdeps/ia64/backtrace.c:85
#7  0x2aae591c in __mf_backtrace (symbols=0x7ffdf448, guess_pc=0x2aae7918,
guess_omit_levels=2) at
/home/apinski/src/gcc-fsf/local/gcc/libmudflap/mf-runtime.c:1981
#8  0x2aae7088 in __mf_violation (ptr=0x100a1338, sz=0, pc=716077336,
location=<optimized out>, type=4)
    at /home/apinski/src/gcc-fsf/local/gcc/libmudflap/mf-runtime.c:2106
#9  0x2aae74a4 in __mfu_unregister (ptr=0x100a1338, sz=0, type=<optimized out>)
at /home/apinski/src/gcc-fsf/local/gcc/libmudflap/mf-runtime.c:1351
#10 0x2aae7918 in __mf_unregister (ptr=0x100a1338, sz=0, type=2) at
/home/apinski/src/gcc-fsf/local/gcc/libmudflap/mf-runtime.c:1291
#11 0x2aae95c4 in free (buf=0x100a1338) at
/home/apinski/src/gcc-fsf/local/gcc/libmudflap/mf-hooks1.c:268
#12 0x2ad40f10 in __deregister_frame_info_bases (begin=0x2ac04ebc) at
../sysdeps/generic/unwind-dw2-fde.c:227
#13 0x2abf4164 in _ftext () from /lib32/octeon2/libpthread.so.0
#14 0x2ac03ebc in _fini () from /lib32/octeon2/libpthread.so.0
#15 0x2aabfb20 in _dl_fini () at dl-fini.c:248
#16 0x2ac53f70 in __GI_exit (status=0) at exit.c:75
#17 0x2ac38ba8 in __libc_start_main (main=0x10000a30 <__wrap_main@plt>, argc=1,
ubp_av=0x7ffdf804, init=<optimized out>, fini=<optimized out>,
rtld_fini=<optimized out>, 
    stack_end=<optimized out>) at libc-start.c:252
#18 0x10001020 in __start ()

--- CUT ---
__deregister_frame_info_bases has uses a mutex which is also used from
_Unwind_Find_registered_FDE  which is what is causing the infinite loop. 
Though we should not have an unregisterd memory location here though.