Wireshark-dev: Re: [Wireshark-dev] g_mem-chunk_destroy does not free memory for GLib > 2.8
From: Bill Meier <[email protected]>
Date: Tue, 20 Oct 2009 18:26:13 -0400
Jeff Morriss wrote:
Bill Meier wrote:
Jeff Morriss wrote:
Bill Meier wrote:
   I've found at least one case wherein se_alloc'd memory
   is referenced after it's been freed. (I think this case just happens
   to work because the memory wasn't yet reassigned).
I was contemplating adding an optional memory-scrubber to ensure that we're not doing that: just writing 0xdeadbeef to the ep_ and se_ alloc'd memory when we free it.
I was thinking the same thing ....
I fixed a few cases of this and I'm not getting any more failures with 
the sample captures I have.  Is the case you found fixed?
Unfortunately I don't quite remember the case I found.

Was it the "dangling pointer to se'alloc'd memory" I mentioned in bug #4141 ? (If so I originally mis-interpreted the code in that there's no actual reference to the already freed se_alloc'd buffer).
      "Also note that currently for free_all_reassembled_fragments()
      that the 'key_arg' points to se_alloc'd memory which has already
      been freed."

> If so, should we turn on memory scrubbing while fuzz testing?

Seems like a reasonable idea to me. We'd need to watch to see how much time is added to the fuzz-testing. Maybe only turn on scrubbing for one pass ?