[ruby-core:120130] [Ruby master Bug#20934] `UnboundMethod#bind_call` may cause "double free or corruption" with Ractor

Issue #20934 has been updated by wanabe (_ wanabe).


Thanks for your comment. It looks like a race condition between `rb_method_definition_release()` and `method_definition_addref()`.
Applying the following patch suppressed the problem.

```patch
diff --git a/vm_method.c b/vm_method.c
index 69a533b182..725b2a99ac 100644
--- a/vm_method.c
+++ b/vm_method.c
@@ -518,6 +518,7 @@ static void
 rb_method_definition_release(rb_method_definition_t *def)
 {
     if (def != NULL) {
+        RB_VM_LOCK_ENTER();
         const int reference_count = def->reference_count;
         def->reference_count--;
 
@@ -535,6 +536,7 @@ rb_method_definition_release(rb_method_definition_t *def)
             if (METHOD_DEBUG) fprintf(stderr, "-%p-%s:%d->%d (dec)\n", (void *)def, rb_id2name(def->original_id),
                                       reference_count, def->reference_count);
         }
+        RB_VM_LOCK_LEAVE();
     }
 }
 
@@ -621,9 +623,11 @@ setup_method_cfunc_struct(rb_method_cfunc_t *cfunc, VALUE (*func)(ANYARGS), int
 static rb_method_definition_t *
 method_definition_addref(rb_method_definition_t *def, bool complemented)
 {
+    RB_VM_LOCK_ENTER();
     if (!complemented &&  def->reference_count > 0) def->aliased = true;
     def->reference_count++;
     if (METHOD_DEBUG) fprintf(stderr, "+%p-%s:%d\n", (void *)def, rb_id2name(def->original_id), def->reference_count);
+    RB_VM_LOCK_LEAVE();
     return def;
 }
 
```

But I don't think it is a good idea because the cost would be high.
Hopefully there is a better solution.

----------------------------------------
Bug #20934: `UnboundMethod#bind_call` may cause "double free or corruption" with Ractor
https://bugs.ruby-lang.org/issues/20934#change-110880

* Author: wanabe (_ wanabe)
* Status: Open
* ruby -v: ruby 3.4.0dev (2024-12-06T18:51:08Z :detached: 8ad6860ff7) +PRISM [x86_64-linux]
* Backport: 3.1: UNKNOWN, 3.2: UNKNOWN, 3.3: UNKNOWN
----------------------------------------
When I call `UnboundMethod#bind_call` on both main Ractor and child Ractor, probable errors can be encountered.
Here is an issue reproduce script `ractor_issue.rb`.
```
def foo
  10000.times do
    Object.instance_method(:object_id).bind_call(self)
  end
end

Ractor.new { foo }
foo
```

And there are some examples of execution results.

```
$ ./miniruby -v ractor_issue.rb 
ruby 3.4.0dev (2024-12-06T18:51:08Z :detached: 8ad6860ff7) +PRISM [x86_64-linux]
ractor_issue.rb:7: warning: Ractor is experimental, and the behavior may change in future versions of Ruby! Also there are many implementation issues.
double free or corruption (fasttop)
Aborted (core dumped)
```

```
$ ./miniruby -v ractor_issue.rb 
ruby 3.4.0dev (2024-12-06T18:51:08Z :detached: 8ad6860ff7) +PRISM [x86_64-linux]
ractor_issue.rb:7: warning: Ractor is experimental, and the behavior may change in future versions of Ruby! Also there are many implementation issues.
#<Thread:0x00007fc6c16bfb18 run> terminated with exception (report_on_exception is true):
ractor_issue.rb:3:in 'UnboundMethod#bind_call': undefined method 'object_id' for main (NoMethodError)
        from ractor_issue.rb:3:in 'block in Object#foo'
        from <internal:numeric>:257:in 'Integer#times'
        from ractor_issue.rb:2:in 'Object#foo'
        from ractor_issue.rb:8:in '<main>'
```

Please try running it several times, as there is a probability of successful completion.



-- 
https://bugs.ruby-lang.org/
 ______________________________________________
 ruby-core mailing list -- [e-mail]
 To unsubscribe send an email to [e-mail]
 ruby-core info -- https://ml.ruby-lang.org/mailman3/lists/ruby-core.ml.ruby-lang.org/