Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 5 Oct 2003 22:32:00 -0500
From:      Dan Nelson <dnelson@allantgroup.com>
To:        Robert Huff <roberthuff@rcn.com>
Cc:        questions@freebsd.org
Subject:   Re: malloc() behavior (was: Pointer please)
Message-ID:  <20031006033200.GL5283@dan.emsphone.com>
In-Reply-To: <16256.57227.924291.290786@jerusalem.litteratus.org>
References:  <27DDB356-F790-11D7-9174-003065838A88@mulle-kybernetik.com> <20031006030656.GK5283@dan.emsphone.com> <16256.57227.924291.290786@jerusalem.litteratus.org>

next in thread | previous in thread | raw e-mail | index | archive | help
In the last episode (Oct 05), Robert Huff said:
> Dan Nelson writes:
> >  Could be one of two problems.  The program either malloced memory
> >  and tried to use it without zeroing it, or it freed some memory
> >  and tried to keep using it.  In -current, the malloc has the J
> >  debugging flag set, which fills malloced and freed memory with
> >  0xd0 (see the malloc manpage).
> 
> 	On that page (on my 5.1 system), it says malloc() does not zero
> allocated pages.  Is this a change (possibly just for CURRENT), and
> if so since when?  Bexause unless I'm delusional (possible) I thought
> pages /were/ supposed to be zeroed, and doing so was one of the
> system's "as time permits" chores.

Pages handed to processes by the kernel are always zeroed, but pages
free()d then malloc()ed again are not zeroed by default on -RELEASEs,
because they usually aren't returned back to the kernel inbetween
(unless H is set, and even then it's not guaranteed).  -CURRENT always
has the J flag set, which means that any memory returned by malloc or
passed to free will get overwritten with 0xD0, to aid debugging. 
That's not mentioned in the manpage, although I think it is mentioned
someplace else (either FAQ or handbook).

-- 
	Dan Nelson
	dnelson@allantgroup.com



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20031006033200.GL5283>