Issue #8978 has been updated by Aaron Patterson.


I took the mprotect example from the ffi man pages.  Seems there must be a bug in the example code.  Anyway, I've fixed it.

Also, we *should* be using `ffi_prep_closure_loc` if it is available:

  https://github.com/ruby/ruby/blob/92a7f323678d6070383364f2129a5202b217c5da/ext/fiddle/closure.c#L231-234

----------------------------------------
Bug #8978: Fiddle possibly misuses mprotect
https://bugs.ruby-lang.org/issues/8978#change-44663

* Author: Yusuke Endoh
* Status: Closed
* Priority: Normal
* Assignee: Aaron Patterson
* Category: ext
* Target version: 2.1.0
* ruby -v: ruby 2.1.0dev (2013-10-02 trunk 43121) [x86_64-linux]
* Backport: 1.9.3: UNKNOWN, 2.0.0: UNKNOWN
----------------------------------------
Hello Aaron,

Coverity Scan found a possible bug in "initialize" function of ext/fiddle/closure.c:

> result = ffi_prep_closure(pcl, cif, callback, (void *)self);
> ...
> i = mprotect(pcl, sizeof(pcl), PROT_READ | PROT_EXEC)

I don't understand the code completely, but the size of the pointer does not seem to make sense.
Perhaps, "sizeof(pcl)" should be "sizeof(*pcl)".

The same applies to dealloc:

> munmap(cls->pc1, sizeof(cls->pc1));


BTW, ffi_prep_closure seems deprecated.
We should use ffi_prep_closure_loc instead when it is available.

-- 
Yusuke Endoh <mame / tsg.ne.jp>



-- 
http://bugs.ruby-lang.org/