[Mono-bugs] [Bug 76331][Cri] New - Objects on the stack can have
their contents mangled
bugzilla-daemon at bugzilla.ximian.com
bugzilla-daemon at bugzilla.ximian.com
Tue Oct 4 18:08:26 EDT 2005
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 will at cambia.com.
http://bugzilla.ximian.com/show_bug.cgi?id=76331
--- shadow/76331 2005-10-04 18:08:26.000000000 -0400
+++ shadow/76331.tmp.21585 2005-10-04 18:08:26.000000000 -0400
@@ -0,0 +1,145 @@
+Bug#: 76331
+Product: Mono: Runtime
+Version: 1.1
+OS:
+OS Details: Fedora Core 4, Red Hat Enterprise 4
+Status: NEW
+Resolution:
+Severity:
+Priority: Critical
+Component: GC
+AssignedTo: lupus at ximian.com
+ReportedBy: will at cambia.com
+QAContact: mono-bugs at ximian.com
+TargetMilestone: ---
+URL:
+Cc:
+Summary: Objects on the stack can have their contents mangled
+
+Description of Problem:
+
+Under certain circumstances, it appears as if a structure on the stack
+which contains a reference to a heap object has its contents mangled.
+
+Steps to reproduce the problem:
+
+Compile & run the following program:
+
+using System;
+using System.IO;
+using System.Reflection;
+
+namespace t {
+ public struct alloc {
+ public byte b1; public byte b2; public byte b3; public
+byte b4;
+ public byte b5; public byte b6; public byte b7; public
+byte b8;
+ public substruct sub;
+ public byte b13; public byte b14; public byte b15; public
+byte b16;
+ public byte b17; public byte b18; public byte b19; public
+byte b20;
+ public byte b21; public byte b22; public byte b23; public
+byte b24;
+ public byte b25; public byte b26; public byte b27; public
+byte b28;
+ public byte b29; public byte b30; public byte b31; public
+byte b32;
+
+ public void DebugPrint() {
+ string[] s = new string[sub.arr.Length];
+ for (int i = 0; i < s.Length; i++) s[i] = sub.arr
+[i].ToString("x2");
+ Console.WriteLine("{0}", string.Join(" ", s));
+ }
+ }
+ public struct substruct {
+ public byte b1;
+ public byte b2;
+ public byte[] arr;
+ }
+
+ class Test {
+ public static object Alloc(int num) {
+ object[] ret = new object[10 * num];
+ for (int i = 0; i < 10 * num; i++)
+ ret[i] = new string((char)i, 50);
+ return ret;
+ }
+ public static object Create(Type _type) {
+ object o = Alloc(1);
+ object ret = Activator.CreateInstance(_type);
+
+ FieldInfo[] fields = _type.GetFields();
+ foreach (FieldInfo field in fields) {
+ Type fieldtype = field.FieldType;
+ object o2 = Alloc(1);
+ if (fieldtype.Equals(typeof(byte))) {
+ object o3 = Alloc(3);
+ field.SetValue(ret,
+Activator.CreateInstance(fieldtype));
+ } else if (fieldtype.Equals(typeof
+(substruct))) {
+ object o3 = Alloc(1);
+ field.SetValue(ret, Create
+(fieldtype));
+ } else {
+ object o3 = Alloc(2);
+ field.SetValue(ret, new byte[4]);
+ }
+ }
+ return ret;
+ }
+ static void Main() {
+ alloc t = (alloc)Create(typeof(alloc));
+
+ while (t.sub.arr.Length == 4) {
+ t.DebugPrint();
+ }
+ }
+ }
+}
+
+Actual Results:
+
+The program either throws an exception or exits
+
+Expected Results:
+
+The program should run forever
+
+How often does this happen?
+
+Every time.
+
+Additional Information:
+
+What this program does is create a specially-laid-out structure in a
+specific way, allocating heap memory as it goes. The reproduction of
+this bug is very dependant on the amount of memory you allocate while
+constructing the object in question. After the object is constructed and
+placed on the stack in Main(), it will contain a reference to an array of
+4 bytes. The program then enters a while loop and prints out the
+contents of the array each time through. After some period of time (it
+happens asynchronously but this test program it always happens very
+quickly for me) the array reference will no longer be valid. The address
+stored in the structure has stayed the same but it appears the garbage
+collector has ignored the outstanding reference and collected the memory
+it points to. This causes the array reference to point to random garbage
+(this typically causes the loop to throw a NullReferenceException, but it
+can also change the size or contents of the array as well). Oddly enough
+if I add a GC.Collect() directly after the allocation this will prevent
+the bug from occurring. Apparently there is a difference between the
+full collection and whatever it is doing asynchronously when GC.Collect
+is not present.
+
+While finding a small example which could reproduce this problem was
+difficult, this behavior occurs regularly in our production code and is
+causing immense havoc, since objects can no longer be trusted to contain
+the values they were set to. Any help is greatly appreciated.
+
+The above code reproduces the bug on a celeron running Red Hat Enterprise
+4 as well as a dual Xeon running Fedora Core 4. I don't know if it will
+work on other machines as it seems highly dependant on the behavior of
+the memory allocator.
More information about the mono-bugs
mailing list