Re: [PATCH] Reclaim _REENT_MP_P5S in _reclaim_reent

On Nov 16 12:12, [email protected] wrote:
> From: Chris Johns <[email protected]>
> 
> The change fixes a memory leak in threads that clean up using
> _reclaim_reent.
> 
> RTEMS: Closes #4967
> ---
>  newlib/libc/reent/reent.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/newlib/libc/reent/reent.c b/newlib/libc/reent/reent.c
> index db80ca06e..6bb271b9a 100644
> --- a/newlib/libc/reent/reent.c
> +++ b/newlib/libc/reent/reent.c
> @@ -59,6 +59,17 @@ _reclaim_reent (struct _reent *ptr)
>  	}
>        if (_REENT_MP_RESULT(ptr))
>  	_free_r (ptr, _REENT_MP_RESULT(ptr));
> +      if (_REENT_MP_P5S(ptr))
> +        {
> +         struct _Bigint *thisone, *nextone;
> +         nextone = _REENT_MP_P5S(ptr);
> +         while (nextone)
> +           {
> +             thisone = nextone;
> +             nextone = nextone->_next;
> +             _free_r (ptr, thisone);
> +           }
> +       }

The p5s data is allocated using Balloc, so the pointers are in freelist
and should already be free'd at this point, starting at line 42.

Please explain what happens and why free'ing the freelist is not
sufficient in this case, preferredly as part of the commit message.


Thanks,
Corinna