Re: [PATCH] staging: comedi: cast to (unsigned int *)

[Greg KH <gregkh@linuxfoundation.org>]

On Wed, Feb 17, 2021 at 11:40:00PM +0530, Atul Gopinathan wrote:
> On Wed, Feb 17, 2021 at 06:35:15PM +0100, Greg KH wrote:
> > On Wed, Feb 17, 2021 at 10:29:08PM +0530, Atul Gopinathan wrote:
> > > Resolve the following warning generated by sparse:
> > > 
> > > drivers/staging//comedi/comedi_fops.c:2956:23: warning: incorrect type in assignment (different address spaces)
> > > drivers/staging//comedi/comedi_fops.c:2956:23:    expected unsigned int *chanlist
> > > drivers/staging//comedi/comedi_fops.c:2956:23:    got void [noderef] <asn:1> *
> > > 
> > > compat_ptr() has a return type of "void __user *"
> > > as defined in "include/linux/compat.h"
> > > 
> > > cmd->chanlist is of type "unsigned int *" as defined
> > > in drivers/staging/comedi/comedi.h" in struct
> > > comedi_cmd.
> > > 
> > > Signed-off-by: Atul Gopinathan <atulgopinathan@gmail.com>
> > > ---
> > >  drivers/staging/comedi/comedi_fops.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/drivers/staging/comedi/comedi_fops.c b/drivers/staging/comedi/comedi_fops.c
> > > index e85a99b68f31..fc4ec38012b4 100644
> > > --- a/drivers/staging/comedi/comedi_fops.c
> > > +++ b/drivers/staging/comedi/comedi_fops.c
> > > @@ -2953,7 +2953,7 @@ static int get_compat_cmd(struct comedi_cmd *cmd,
> > >  	cmd->scan_end_arg = v32.scan_end_arg;
> > >  	cmd->stop_src = v32.stop_src;
> > >  	cmd->stop_arg = v32.stop_arg;
> > > -	cmd->chanlist = compat_ptr(v32.chanlist);
> > > +	cmd->chanlist = (unsigned int __force *)compat_ptr(v32.chanlist);
> > 
> > __force?  That feels wrong, something is odd if that is ever needed.
> > 
> > Are you _sure_ this is correct?
> 
> The same file has instances of "(usigned int __force *)" cast being
> used on the same "cmd->chanlist". For reference:
> 
> At line 1797 of comedi_fops.c:
> 1796                 /* restore chanlist pointer before copying back */
> 1797                 cmd->chanlist = (unsigned int __force *)user_chanlist;
> 1798                 cmd->data = NULL;
> 
> At line 1880:
> 1879         /* restore chanlist pointer before copying back */
> 1880         cmd->chanlist = (unsigned int __force *)user_chanlist;
> 1881         *copy = true;
> 
> Here "user_chanlist" is of type "unsigned int __user *".
> 
> 
> Or perhaps, I shouldn't be relying on them?

I don't know, it still feels wrong.

Ian, any thoughts?

thanks,

greg k-h