py: Fix VM crash with unwinding jump out of a finally block.

dpgeorge · dpgeorge · commit e1fb03f3e2d1 · 2019-03-05T16:05:05.000+11:00
This patch fixes a bug in the VM when breaking within a try-finally. The bug has to do with executing a break within the finally block of a try-finally statement. For example: def f(): for x in (1,): print('a', x) try: raise Exception finally: print(1) break print('b', x) f() Currently in uPy the above code will print: a 1 1 1 segmentation fault (core dumped) micropython Not only is there a seg fault, but the "1" in the finally block is printed twice. This is because when the VM executes a finally block it doesn't really know if that block was executed due to a fall-through of the try (no exception raised), or because an exception is active. In particular, for nested finallys the VM has no idea which of the nested ones have active exceptions and which are just fall-throughs. So when a break (or continue) is executed it tries to unwind all of the finallys, when in fact only some may be active. It's questionable whether break (or return or continue) should be allowed within a finally block, because they implicitly swallow any active exception, but nevertheless it's allowed by CPython (although almost never used in the standard library). And uPy should at least not crash in such a case. The solution here relies on the fact that exception and finally handlers always appear in the bytecode after the try body. Note: there was a similar bug with a return in a finally block, but that was previously fixed in b735208
diff --git a/py/compile.c b/py/compile.c
@@ -1599,7 +1599,6 @@ STATIC void compile_try_except(compiler_t *comp, mp_parse_node_t pn_body, int n_
         }
         compile_node(comp, pns_except->nodes[1]); // the <body>
         if (qstr_exception_local != 0) {
-            EMIT(pop_block);
             EMIT_ARG(load_const_tok, MP_TOKEN_KW_NONE);
             EMIT_ARG(label_assign, l3);
             EMIT_ARG(load_const_tok, MP_TOKEN_KW_NONE);
@@ -1635,7 +1634,6 @@ STATIC void compile_try_finally(compiler_t *comp, mp_parse_node_t pn_body, int n
     } else {
         compile_try_except(comp, pn_body, n_except, pn_except, pn_else);
     }
-    EMIT(pop_block);
     EMIT_ARG(load_const_tok, MP_TOKEN_KW_NONE);
     EMIT_ARG(label_assign, l_finally_block);
     compile_node(comp, pn_finally);
@@ -1811,8 +1809,7 @@ STATIC void compile_async_with_stmt_helper(compiler_t *comp, int n, mp_parse_nod
         compile_async_with_stmt_helper(comp, n - 1, nodes + 1, body);
         EMIT_ARG(adjust_stack_size, -3);
 
-        // Finish the "try" block
-        EMIT(pop_block);
+        // We have now finished the "try" block and fall through to the "finally"
 
         // At this point, after the with body has executed, we have 3 cases:
         // 1. no exception, we just fall through to this point; stack: (..., ctx_mgr)
diff --git a/py/emitbc.c b/py/emitbc.c
@@ -738,7 +738,6 @@ void mp_emit_bc_setup_block(emit_t *emit, mp_uint_t label, int kind) {
 }
 
 void mp_emit_bc_with_cleanup(emit_t *emit, mp_uint_t label) {
-    mp_emit_bc_pop_block(emit);
     mp_emit_bc_load_const_tok(emit, MP_TOKEN_KW_NONE);
     mp_emit_bc_label_assign(emit, label);
     emit_bc_pre(emit, 2); // ensure we have enough stack space to call the __exit__ method
diff --git a/py/vm.c b/py/vm.c
@@ -633,8 +633,6 @@ run_code_state: ;
                             // replacing it with None, which signals END_FINALLY to just
                             // execute the finally handler normally.
                             SET_TOP(mp_const_none);
-                            assert(exc_sp >= exc_stack);
-                            POP_EXC_BLOCK();
                         } else {
                             // We need to re-raise the exception.  We pop __exit__ handler
                             // by copying the exception instance down to the new top-of-stack.
@@ -654,7 +652,7 @@ unwind_jump:;
                     while ((unum & 0x7f) > 0) {
                         unum -= 1;
                         assert(exc_sp >= exc_stack);
-                        if (MP_TAGPTR_TAG1(exc_sp->val_sp)) {
+                        if (MP_TAGPTR_TAG1(exc_sp->val_sp) && exc_sp->handler > ip) {
                             // Getting here the stack looks like:
                             //     (..., X, dest_ip)
                             // where X is pointed to by exc_sp->val_sp and in the case
@@ -698,6 +696,8 @@ unwind_jump:;
                     // if TOS is an integer, finishes coroutine and returns control to caller
                     // if TOS is an exception, reraises the exception
                     if (TOP() == mp_const_none) {
+                        assert(exc_sp >= exc_stack);
+                        POP_EXC_BLOCK();
                         sp--;
                     } else if (mp_obj_is_small_int(TOP())) {
                         // We finished "finally" coroutine and now dispatch back
@@ -1063,7 +1063,7 @@ unwind_jump:;
 unwind_return:
                     // Search for and execute finally handlers that aren't already active
                     while (exc_sp >= exc_stack) {
-                        if (!currently_in_except_block && MP_TAGPTR_TAG1(exc_sp->val_sp)) {
+                        if (!currently_in_except_block && MP_TAGPTR_TAG1(exc_sp->val_sp) && exc_sp->handler > ip) {
                             // Found a finally handler that isn't active.
                             // Getting here the stack looks like:
                             //     (..., X, [iter0, iter1, ...,] ret_val)
@@ -1419,7 +1419,7 @@ unwind_jump:;
                 mp_obj_exception_add_traceback(MP_OBJ_FROM_PTR(nlr.ret_val), source_file, source_line, block_name);
             }
 
-            while (currently_in_except_block) {
+            while (currently_in_except_block || (exc_sp >= exc_stack && exc_sp->handler <= code_state->ip)) {
                 // nested exception
 
                 assert(exc_sp >= exc_stack);
diff --git a/tests/basics/try_finally_break.py b/tests/basics/try_finally_break.py
@@ -0,0 +1,99 @@
+# test break within (nested) finally
+
+# basic case with break in finally
+def f():
+    for _ in range(2):
+        print(1)
+        try:
+            pass
+        finally:
+            print(2)
+            break
+            print(3)
+        print(4)
+    print(5)
+f()
+
+# where the finally swallows an exception
+def f():
+    lst = [1, 2, 3]
+    for x in lst:
+        print('a', x)
+        try:
+            raise Exception
+        finally:
+            print(1)
+            break
+        print('b', x)
+f()
+
+# basic nested finally with break in inner finally
+def f():
+    for i in range(2):
+        print('iter', i)
+        try:
+            raise TypeError
+        finally:
+            print(1)
+            try:
+                raise ValueError
+            finally:
+                break
+print(f())
+
+# similar to above but more nesting
+def f():
+    for i in range(2):
+        try:
+            raise ValueError
+        finally:
+            print(1)
+            try:
+                raise TypeError
+            finally:
+                print(2)
+                try:
+                    pass
+                finally:
+                    break
+print(f())
+
+# lots of nesting
+def f():
+    for i in range(2):
+        try:
+            raise ValueError
+        finally:
+            print(1)
+            try:
+                raise TypeError
+            finally:
+                print(2)
+                try:
+                    raise Exception
+                finally:
+                    break
+print(f())
+
+# basic case combined with try-else
+def f(arg):
+    for _ in range(2):
+        print(1)
+        try:
+            if arg == 1:
+                raise ValueError
+            elif arg == 2:
+                raise TypeError
+        except ValueError:
+            print(2)
+        else:
+            print(3)
+        finally:
+            print(4)
+            break
+            print(5)
+        print(6)
+    print(7)
+f(0) # no exception, else should execute
+f(1) # exception caught, else should be skipped
+f(2) # exception not caught, finally swallows exception, else should be skipped
diff --git a/tests/basics/try_return.py b/tests/basics/try_return.py
@@ -1,5 +1,14 @@
 # test use of return with try-except
 
+def f():
+    try:
+        print(1)
+        return
+    except:
+        print(2)
+    print(3)
+f()
+
 def f(l, i):
     try:
         return l[i]
diff --git a/tests/cmdline/cmd_showbc.py.exp b/tests/cmdline/cmd_showbc.py.exp
@@ -265,7 +265,6 @@ Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
 \\d\+ POP_EXCEPT
 \\d\+ JUMP \\d\+
 \\d\+ END_FINALLY
-\\d\+ POP_BLOCK
 \\d\+ LOAD_CONST_NONE
 \\d\+ LOAD_FAST 1
 \\d\+ POP_TOP
@@ -286,7 +285,6 @@ Raw bytecode (code_info_size=\\d\+, bytecode_size=\\d\+):
 \\d\+ POP_TOP
 \\d\+ LOAD_DEREF 14
 \\d\+ POP_TOP
-\\d\+ POP_BLOCK
 \\d\+ LOAD_CONST_NONE
 \\d\+ WITH_CLEANUP
 \\d\+ END_FINALLY

Original file line number	Diff line number	Diff line change
`@@ -738,7 +738,6 @@ void mp_emit_bc_setup_block(emit_t *emit, mp_uint_t label, int kind) {`
`738`	`738`	`}`
`739`	`739`
`740`	`740`	`void mp_emit_bc_with_cleanup(emit_t *emit, mp_uint_t label) {`
`741`		`- mp_emit_bc_pop_block(emit);`
`742`	`741`	`mp_emit_bc_load_const_tok(emit, MP_TOKEN_KW_NONE);`
`743`	`742`	`mp_emit_bc_label_assign(emit, label);`
`744`	`743`	`emit_bc_pre(emit, 2); // ensure we have enough stack space to call the __exit__ method`