Re: [PATCH][gdb] Fix heap-buffer-overflow in completion_tracker::build_completion_result

[Pedro Alves <pedro@palves.net>]

On 12/4/20 5:39 PM, Tom de Vries wrote:
> Hi,
> 
> When building gdb with address sanitizer and running test-case
> gdb.base/completion.exp, we run into:
> ...
> ==5743==ERROR: AddressSanitizer: heap-buffer-overflow on address \
>   0x60200025c02f at pc 0x000000cd9d64 bp 0x7fff3297da30 sp 0x7fff3297da28
> READ of size 1 at 0x60200025c02f thread T0
>     #0 0xcd9d63 in completion_tracker::build_completion_result(char const*, \
>                      int, int) gdb/completer.c:2258
>   ...
> 0x60200025c02f is located 1 bytes to the left of 1-byte region \
>   [0x60200025c030,0x60200025c031)
> ...
> 

It would be useful to show which test triggered this.  I checked, and it's
the new:

 (gdb) p/d[TAB]

test.

> The problem is in this code in completion_tracker::build_completion_result:
> ...
>       bool completion_suppress_append
>         = (suppress_append_ws ()
>            || match_list[0][strlen (match_list[0]) - 1] == ' ');
> ...
> If strlen (match_list[0]) == 0, then we access match_list[0][-1].
> 
> Fix this by testing if the memory access is in bounds before doing the memory
> access.
> 
> Tested on x86_64-linux.
> 
> Any comments?

My first question is -- why do we end up here in the first place?
Why is m_lowest_common_denominator_unique set with the lcd empty?

The answer is that the new "/FMT" completer pushes an empty
completion match:

	  if (*text == '\0')
	    {
	      /* We're at the end of the input string.  The user has typed
		 '/FMT' and asked for a completion.  Push an empty
		 completion string, this will cause readline to insert a
		 space so the user now has '/FMT '.  */
	      in_fmt = true;
	      tracker.add_completion (make_unique_xstrdup (text));
	    }

This could also be fixed by adding a non-empty completion and not advancing
the completion word point, like:

 -             tracker.add_completion (make_unique_xstrdup (text));
 +             tracker.add_completion (make_unique_xstrdup (*args));
 +             return true;

but I'm not thinking of a real problem the empty completion match
causes, so I think it's OK to properly support it.

So the patch is OK.  But please mention the "p/d[TAB]" scenario in
the commit log.