cvs commit: src/sys/fs/devfs devfs_vnops.c src/sys/kern kern_conf.c src/sys/sys conf.h

John Baldwin jhb at FreeBSD.org
Wed Aug 17 17:28:21 GMT 2005


On Wednesday 17 August 2005 01:04 pm, Nate Lawson wrote:
> Poul-Henning Kamp wrote:
> > phk         2005-08-17 08:19:53 UTC
> >
> >   FreeBSD src repository
> >
> >   Modified files:
> >     sys/fs/devfs         devfs_vnops.c
> >     sys/kern             kern_conf.c
> >     sys/sys              conf.h
> >   Log:
> >   Handle device drivers with D_NEEDGIANT in a way which does not
> >   penalize the 'good' drivers:  Allocate a shadow cdevsw and populate
> >   it with wrapper functions which grab Giant
> >
> >   Revision  Changes    Path
> >   1.120     +0 -24     src/sys/fs/devfs/devfs_vnops.c
> >   1.190     +155 -9    src/sys/kern/kern_conf.c
> >   1.226     +1 -0      src/sys/sys/conf.h
> >
> >
> >  static void
> >  prep_cdevsw(struct cdevsw *devsw)
> >  {
> > +	struct cdevsw *dsw2;
> >
> > +	if (devsw->d_flags & D_NEEDGIANT)
> > +		dsw2 = malloc(sizeof *dsw2, M_DEVT, M_WAITOK);
> > +	else
> > +		dsw2 = NULL;
> >  	dev_lock();
> >
> >  	if (devsw->d_version != D_VERSION_01) {
>
> Is it guaranteed that prep_cdevsw() won't be called twice without a call
> to fini_cdevsw()?  Otherwise, the shadow cdevsw could leak when its
> pointer is set to NULL.  I'm not familiar with this code, just asking.

Err, how exactly?  If devsw->d_gianttrick is already non-NULL, it just free's 
dsw2.

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org


More information about the cvs-src mailing list