Re: 9pfs double kfree

[Pavel Machek <pavel@ucw.cz>]

On Po 06-03-06 09:34:01, Al Viro wrote:
> On Mon, Mar 06, 2006 at 10:40:03AM +0200, Kai Makisara wrote:
> > > Legal, but rather bad taste.  Init to NULL, possibly assign the value
> > > if kmalloc(), then kfree() unconditionally - sure, but that... almost
> > > certainly one hell of a lousy cleanup logics somewhere.
> > > 
> > I agree with you.
> > 
> > However, a few months ago it was advocated to let kfree take care of 
> > testing the pointer against NULL and a load of patches like this:
> 
> That's different - that's _exactly_ the case I've mentioned above.
> 
> Moreover, that's exact match to standard behaviour of free(3):
> 
> C99 7.20.3.2(2):
> The free function causes the space pointed to by ptr to be deallocated, that
> is, made available for further allocation.  If ptr is a null pointer, no action
> occurs.  Otherwise, if the argument does not match a pointer returned by the
> calloc, malloc, or realloc function, or if the space has been deallocated by
> a call to free or realloc, the behaviour is undefined.
> 
> IOW, free(NULL) is guaranteed to be no-op while double-free is nasal daemon
> country.

Well, double-free of NULL is permitted by text above. 'If ptr is a
null pointer, no action occurs.'

OTOH #define kfree(a) { __kfree(a); a = NULL; } actually does the
right thing... even with double free.
								Pavel
-- 
Web maintainer for suspend.sf.net (www.sf.net/projects/suspend) wanted...