Re: 9pfs double kfree

[Dave Jones <davej@redhat.com>]

On Mon, Mar 06, 2006 at 12:56:03PM +0530, Balbir Singh wrote:
 > On 3/6/06, David S. Miller <davem@davemloft.net> wrote:
 > > From: Dave Jones <davej@redhat.com>
 > > Date: Mon, 6 Mar 2006 02:04:58 -0500
 > >
 > > > (I wish we had a kfree variant that NULL'd the target when it was free'd)
 > >
 > > Excellent idea.
 > > -
 > 
 > Slab debugging should catch double frees, but it will not attract your
 > attention till you see your dmesg log.

The vast majority of people never run with slab poisoning enabled
judging by the number of bugs it constantly turns up.

 > kfree() will ignore NULL
 > pointer, from the comments in kfree

*nod*, poisoning the ptr would be a better idea.

 > May we could have such a variant under CONFIG_DEBUG_SLAB if we needed
 > and also change the variant kfree to BUG_ON() a NULL pointer.

given the cost is just a ptr assignment in a slow path, I'd prefer
it was non-optional, otherwise it'll be as underused as the other
debugging options.

		Dave

-- 
http://www.codemonkey.org.uk