Re: [RFC] [PATCH] Pre-emption control for userspace

From: Al Viro <viro@ZenIV.linux.org.uk>
To: Khalid Aziz <khalid.aziz@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>,
	tglx@linutronix.de, mingo@redhat.com, hpa@zytor.com,
	peterz@infradead.org, akpm@linux-foundation.org,
	andi.kleen@intel.com, rob@landley.net, venki@google.com,
	linux-kernel@vger.kernel.org
Subject: Re: [RFC] [PATCH] Pre-emption control for userspace
Date: Tue, 4 Mar 2014 18:38:58 +0000	[thread overview]
Message-ID: <20140304183857.GU18016@ZenIV.linux.org.uk> (raw)
In-Reply-To: <53161116.9050109@oracle.com>

On Tue, Mar 04, 2014 at 10:44:54AM -0700, Khalid Aziz wrote:

> do_exit() unmaps mmap_state->uaddr, and frees up mmap_state->kaddr
> and mmap_state. mmap_state should not be NULL after unmap. vfree()
> and kfree() are tolerant of pointers that have already been freed.

Huh?  Double free() is a bug, plain and simple.  Never do that - not
in userland and especially not in the kernel.  Think what happens if
some code gets executed between those two and asks to allocate something.
If it gets the area you'd just freed, your second free will leave it
with all kinds of nasty surprises.  Starting with "who the hell has
started to modify the object I'd allocated and hadn't freed?"

A:	p = alloc();
A:	free(p);
B:	q = alloc();	/* q == p now */
B:	*q = 0;		/* *q is zero */
A:	free(p);	/* same as free(q) */
C:	r = alloc();	/* r == q now */
C:	*r = 1;		/* *q is one */
B:	if (*q != 0) panic("somebody's buggering my memory");

It's always a bug, whether the implementation catches it or not.