Repair bounds-checking more, this time tcc -b -run tcc.c -run tcc.c -run tcctest.c works

Hello up there. On the list Grischka made a point that we can't recommend using
-b as long as tcc -b tcc.c doesn't produce anything useful. Now it does, so
please don't treat -b mode as second class citizen anymore.

Thanks,
Kirill

* bcheck2:
  tests: Add tests for compile/run tcc.c with `tcc -b` then compile tcc.c again, then run tcctest.c
  lib/bcheck: Fix code typo in __bound_delete_region()
  lib/bcheck: Don't assume heap goes right after bss
  Make tcc work after self-compiling with bounds-check enabled

.gitignore

@@ -11,8 +11,11 @@ tc1.c
 error.c
 i386-gen1.c
 test.out1
+test.out1b
 test.out2
+test.out2b
 test.out3
+test.out3b
 web.sh
 memdebug.c
 bench

lib/bcheck.c

@@ -25,6 +25,7 @@
     && !defined(__DragonFly__) && !defined(__OpenBSD__)
 #include <malloc.h>
 #endif
+#include <unistd.h>
 
 //#define BOUND_DEBUG
 
@@ -94,9 +95,6 @@ static void *saved_realloc_hook;
 static void *saved_memalign_hook;
 #endif
 
-/* linker definitions */
-extern char _end;
-
 /* TCC definitions */
 extern char __bounds_start; /* start of static bounds table */
 /* error message, just for TCC */
@@ -379,9 +377,32 @@ void __bound_init(void)
 
 #if !defined(__TINYC__) && defined(CONFIG_TCC_MALLOC_HOOKS)
     /* malloc zone is also marked invalid. can only use that with
-       hooks because all libs should use the same malloc. The solution
-       would be to build a new malloc for tcc. */
-    start = (unsigned long)&_end;
+     * hooks because all libs should use the same malloc. The solution
+     * would be to build a new malloc for tcc.
+     *
+     * usually heap (= malloc zone) comes right after bss, i.e. after _end, but
+     * not always - either if we are running from under `tcc -b -run`, or if
+     * address space randomization is turned on(a), heap start will be separated
+     * from bss end.
+     *
+     * So sbrk(0) will be a good approximation for start_brk:
+     *
+     *   - if we are a separately compiled program, __bound_init() runs early,
+     *     and sbrk(0) should be equal or very near to start_brk(b) (in case other
+     *     constructors malloc something), or
+     *
+     *   - if we are running from under `tcc -b -run`, sbrk(0) will return
+     *     start of heap portion which is under this program control, and not
+     *     mark as invalid earlier allocated memory.
+     *
+     *
+     * (a) /proc/sys/kernel/randomize_va_space = 2, on Linux;
+     *     usually turned on by default.
+     *
+     * (b) on Linux >= v3.3, the alternative is to read
+     *     start_brk from /proc/self/stat
+     */
+    start = (unsigned long)sbrk(0);
     size = 128 * 0x100000;
     mark_invalid(start, size);
 #endif
@@ -592,7 +613,7 @@ int __bound_delete_region(void *p)
             }
         }
         /* last page */
-        page = get_page(t2_end);
+        page = get_page(t1_end);
         e2 = (BoundEntry *)((char *)page + t2_end);
         for(e=page;e<e2;e++) {
             e->start = 0;

tcc.h

@@ -1127,7 +1127,8 @@ ST_DATA Sym *local_label_stack;
 ST_DATA Sym *global_label_stack;
 ST_DATA Sym *define_stack;
 ST_DATA CType char_pointer_type, func_old_type, int_type, size_type;
-ST_DATA SValue vstack[VSTACK_SIZE], *vtop;
+ST_DATA SValue __vstack[1+/*to make bcheck happy*/ VSTACK_SIZE], *vtop;
+#define vstack  (__vstack + 1)
 ST_DATA int rsym, anon_sym, ind, loc;
 
 ST_DATA int const_wanted; /* true if constant wanted */

tccgen.c

@@ -55,7 +55,7 @@ ST_DATA Sym *define_stack;
 ST_DATA Sym *global_label_stack;
 ST_DATA Sym *local_label_stack;
 
-ST_DATA SValue vstack[VSTACK_SIZE], *vtop;
+ST_DATA SValue __vstack[1+VSTACK_SIZE], *vtop;
 
 ST_DATA int const_wanted; /* true if constant wanted */
 ST_DATA int nocode_wanted; /* true if no code generation wanted for an expression */

tccpp.c

@@ -3041,7 +3041,6 @@ ST_FUNC void preprocess_init(TCCState *s1)
     s1->ifdef_stack_ptr = s1->ifdef_stack;
     file->ifdef_stack_ptr = s1->ifdef_stack_ptr;
 
-    /* XXX: not ANSI compliant: bound checking says error */
     vtop = vstack - 1;
     s1->pack_stack[0] = 0;
     s1->pack_stack_ptr = s1->pack_stack;

tests/Makefile

@@ -12,6 +12,9 @@ TESTS = libtest \
 		test3 \
 		speedtest \
 		btest \
+		test1b\
+		test2b\
+		test3b\
 		weaktest
 
 		# test4 # this test does not seem to work on any platform
@@ -20,6 +23,9 @@ TESTS = libtest \
 # bounds-checking is supported only on i386
 ifneq ($(ARCH),i386)
     TESTS := $(filter-out btest,$(TESTS))
+    TESTS := $(filter-out test1b,$(TESTS))
+    TESTS := $(filter-out test2b,$(TESTS))
+    TESTS := $(filter-out test3b,$(TESTS))
 endif
 
 # these should work too
@@ -87,18 +93,33 @@ test1: test.ref
 	$(TCC) -run tcctest.c > test.out1
 	@if diff -u test.ref test.out1 ; then echo "Auto Test OK"; fi
 
+test1b: test.ref
+	@echo ------------ $@ ------------
+	$(TCC) -b -run tcctest.c > test.out1b
+	@if diff -u test.ref test.out1b ; then echo "Auto Test OK"; fi
+
 # iterated test2 (compile tcc then compile tcctest.c !)
 test2: test.ref
 	@echo ------------ $@ ------------
 	$(TCC) $(RUN_TCC) $(RUN_TCC) -run tcctest.c > test.out2
 	@if diff -u test.ref test.out2 ; then echo "Auto Test2 OK"; fi
 
+test2b: test.ref ../bcheck.o
+	@echo ------------ $@ ------------
+	$(TCC) -b $(RUN_TCC) $(RUN_TCC) -run tcctest.c > test.out2b
+	@if diff -u test.ref test.out2b ; then echo "Auto Test2b OK"; fi
+
 # iterated test3 (compile tcc then compile tcc then compile tcctest.c !)
 test3: test.ref
 	@echo ------------ $@ ------------
 	$(TCC) $(RUN_TCC) $(RUN_TCC) $(RUN_TCC) -run tcctest.c > test.out3
 	@if diff -u test.ref test.out3 ; then echo "Auto Test3 OK"; fi
 
+test3b: test.ref
+	@echo ------------ $@ ------------
+	$(TCC) -b $(RUN_TCC) $(RUN_TCC) $(RUN_TCC) -run tcctest.c > test.out3b
+	@if diff -u test.ref test.out3b ; then echo "Auto Test3 OK"; fi
+
 # binary output test
 test4: test.ref
 	@echo ------------ $@ ------------
@@ -190,5 +211,5 @@ cache: tcc_g
 
 # clean
 clean:
-	rm -vf *~ *.o *.a *.bin *.i *.ref *.out *.out? *.gcc \
+	rm -vf *~ *.o *.a *.bin *.i *.ref *.out *.out? *.out?b *.gcc \
 	   tcctest[1234] ex? libtcc_test$(EXESUF) tcc_g tcclib.h