[Mono-bugs] [Bug 78782][Blo] Changed - Segfault on PPC when using a libgcc_s compiled with SSP

Sun Jul 9 14:50:29 EDT 2006

Please do not reply to this email- if you want to comment on the bug, go to the
URL shown below and enter your comments there.

Changed by slomo at ubuntu.com.

http://bugzilla.ximian.com/show_bug.cgi?id=78782

--- shadow/78782	2006-07-09 11:10:47.000000000 -0400
+++ shadow/78782.tmp.4872	2006-07-09 14:50:29.000000000 -0400
@@ -407,6 +407,145 @@
 #21 0x100d2fa0 in mini_init (filename=0x7fe2fa28
 "../class/lib/monolite/mcs.exe") at mini.c:10492
 #22 0x10014574 in mono_main (argc=13, argv=<value optimized out>) at
 driver.c:867
 #23 0x10013ae0 in main (argc=13, argv=0x7fe2f844) at main.c:12
 
+
+------- Additional Comments From slomo at ubuntu.com  2006-07-09 14:50 -------
+Ok, please ignore everything noted above. As "make clean" doesn't do
+what it should I built everything from a broken source tree and
+additionally the g_malloc -> codeman "fix" only breaks everything harder.
+
+In any case there are still 3 places where malloc() (or realloc()) is
+used for code on PPC:
+- mono_debugger_create_notification_function()
+- mono_arch_emit_prolog()
+- mono_arch_emit_epilog()
+
+
+So here's the current state:
+
+When running mono (1.1.13.8 or 1.1.16) on linux/ppc with glibc 2.4 and
+a libgcc_s.so.1 (from gcc 4.1.1) compiled with SSP mono segfaults
+always. mono works when using a non-SSP libgcc_s.so.1.
+
+Also it works fine on x86, sparc, amd64 and ia64 with the same
+environment.
+
+The segfault happens every time mono is running an assembly, mono
+--help for example works. Building mono obviously doesn't work.
+
+Parameters to configure are --with-gc=included --with-tls=__thread.
+When using --with-gc=none everything works fine. When using
+--with-gc=boehm (using the system libgc that definitely works and
+doesn't crash inkscape for example) one gets the same segfault again.
+Using --with-tls=pthread gives exactly the same results.
+
+Output of mono:
+=================================================================
+Got a SIGSEGV while executing native code. This usually indicates
+a fatal error in the mono runtime or one of the native libraries 
+used by your application.
+=================================================================
+
+Stacktrace:
+
+
+=================================================================
+Got a SIGSEGV while executing native code. This usually indicates
+a fatal error in the mono runtime or one of the native libraries 
+used by your application.
+=================================================================
+
+Stacktrace:
+[and so on]
+
+
+Running it in gdb gives the following insights:
+
+Program received signal SIGSEGV, Segmentation fault.
+0x0fb58604 in _Unwind_FindEnclosingFunction () from /lib/libgcc_s.so.1
+(gdb) bt
+#0  0x0fb58604 in _Unwind_FindEnclosingFunction () from /lib/libgcc_s.so.1
+#1  0x0fb5941c in __frame_state_for () from /lib/libgcc_s.so.1
+#2  0x0fb59f64 in _Unwind_ForcedUnwind () from /lib/libgcc_s.so.1
+#3  0x0fe6c458 in _Unwind_ForcedUnwind () from /lib/libpthread.so.0
+#4  0x0fe696a4 in __pthread_unwind () from /lib/libpthread.so.0
+#5  0x0fe696a4 in __pthread_unwind () from /lib/libpthread.so.0
+#6  0x0fe696a4 in __pthread_unwind () from /lib/libpthread.so.0
+#7  0x0fe696a4 in __pthread_unwind () from /lib/libpthread.so.0
+#8  0x0fe696a4 in __pthread_unwind () from /lib/libpthread.so.0
+#9  0x0fe696a4 in __pthread_unwind () from /lib/libpthread.so.0
+#10 0x0fe696a4 in __pthread_unwind () from /lib/libpthread.so.0
+Previous frame inner to this frame (corrupt stack?)
+
+Now setting a breakpoint on __pthread_unwind() gives the following:
+Breakpoint 2, 0x0fe69668 in __pthread_unwind () from /lib/libpthread.so.0
+(gdb) bt
+#0  0x0fe69668 in __pthread_unwind () from /lib/libpthread.so.0
+#1  0x0fe62e34 in pthread_exit () from /lib/libpthread.so.0
+#2  0x100c6550 in thread_exit (exitstatus=0, handle=0x408) at
+threads.c:146
+#3  0x100c66b8 in thread_start_routine (args=0x30498db8) at threads.c:194
+#4  0x100e1328 in GC_start_routine (arg=<value optimized out>) at
+pthread_support.c:1331
+#5  0x0fe61ab8 in start_thread () from /lib/libpthread.so.0
+#6  0x0fe61ab8 in start_thread () from /lib/libpthread.so.0
+Previous frame inner to this frame (corrupt stack?)
+
+This is the only time __pthread_unwind() is called and doing a "cont"
+gives the same segfault as above again.
+When setting a breakpoint on GC_start_routine() one can see that this
+is the one and only time that it is called.
+
+Running mono with --trace gives the following:
+[...]
+LEAVE: (wrapper runtime-invoke) System.Object:runtime_invoke_void
+(object,intptr,intptr,intptr)[OBJECT:(nil)]
+ENTER: (wrapper runtime-invoke) System.Object:runtime_invoke_void
+(object,intptr,intptr,intptr)() ip: 0x101044c8
+. ENTER: System.IO.StreamWriter:Finalize ()() ip: 0x30c40dd0
+. . ENTER: System.IO.StreamWriter:Dispose (bool)() ip: 0x30c40f08
+. . LEAVE: System.IO.StreamWriter:Dispose (bool)
+. . ENTER: System.Object:Finalize ()() ip: 0x30c40f20
+. . LEAVE: System.Object:Finalize ()
+. LEAVE: System.IO.StreamWriter:Finalize ()
+LEAVE: (wrapper runtime-invoke) System.Object:runtime_invoke_void
+(object,intptr,intptr,intptr)[OBJECT:(nil)]
+ENTER: (wrapper runtime-invoke) System.Object:runtime_invoke_void
+(object,intptr,intptr,intptr)() ip: 0x101044c8
+. ENTER: System.Threading.Thread:Finalize ()() ip: 0x30c40dd0
+. . ENTER: System.IntPtr:op_Explicit (int)() ip: 0x30c41410
+. . . ENTER: System.IntPtr:.ctor (int)() ip: 0x30c4151c
+. . . LEAVE: System.IntPtr:.ctor (int)
+. . LEAVE: System.IntPtr:op_Explicit (int)result=0
+. . ENTER: (wrapper managed-to-native)
+System.Threading.Thread:Thread_free_internal (intptr)() ip: 0x30c41434
+. . LEAVE: (wrapper managed-to-native)
+System.Threading.Thread:Thread_free_internal (intptr)
+. . ENTER: System.Object:Finalize ()() ip: 0x30c4144c
+. . LEAVE: System.Object:Finalize ()
+. LEAVE: System.Threading.Thread:Finalize ()
+LEAVE: (wrapper runtime-invoke) System.Object:runtime_invoke_void
+(object,intptr,intptr,intptr)[OBJECT:(nil)]
+ENTER: (wrapper runtime-invoke) System.Object:runtime_invoke_void
+(object,intptr,intptr,intptr)() ip: 0x101044c8
+. ENTER: System.IO.StreamWrite
+=================================================================
+Got a SIGSEGV while executing native code. This usually indicates
+a fatal error in the mono runtime or one of the native libraries
+used by your application.
+=================================================================
+
+Stacktrace:
+[...]
+
+
+gzipped valgrind output is attached but when comparing the output to
+one generated on x86 this (repeated many times on different addresses)
+looks suspicious:
+Address 0x7EF70768 is just below the stack ptr.  To suppress, use:
+--workaround-gcc296-bugs=yes.
+
+Bye
+
