Re: [PATCH] mei: bus: type promotion bug in mei_nfc_if_version()

[Dan Carpenter <dan.carpenter@oracle.com>]

On Wed, Jul 04, 2018 at 01:57:44PM +0000, Winkler, Tomas wrote:
> > 
> > On Wed, Jul 04, 2018 at 01:59:14PM +0200, Julia Lawall wrote:
> > >
> > >
> > > On Wed, 4 Jul 2018, Dan Carpenter wrote:
> > >
> > > > We accidentally removed the check for negative returns without
> > > > considering the issue of type promotion.  The "if_version_length"
> > > > variable is type size_t so if __mei_cl_recv() returns a negative
> > > > then "bytes_recv" is type promoted to a high positive value and
> > > > treated as success.
> > > >
> > > > Fixes: 582ab27a063a ("mei: bus: fix received data size check in NFC
> > > > fixup")
> > > > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> > > >
> > > > diff --git a/drivers/misc/mei/bus-fixup.c
> > > > b/drivers/misc/mei/bus-fixup.c index 0208c4b027c5..fa0236a5e59a
> > > > 100644
> > > > --- a/drivers/misc/mei/bus-fixup.c
> > > > +++ b/drivers/misc/mei/bus-fixup.c
> > > > @@ -267,7 +267,7 @@ static int mei_nfc_if_version(struct mei_cl *cl,
> > > >
> > > >  	ret = 0;
> > > >  	bytes_recv = __mei_cl_recv(cl, (u8 *)reply, if_version_length, 0);
> > > > -	if (bytes_recv < if_version_length) {
> > > > +	if (bytes_recv < 0 || bytes_recv < if_version_length) {
> > >
> > > Is this preferred to adding an int cast?
> > 
> > I don't think it matters.  I kind of like explicitly testing for negative but
> > maybe later people will just remove the check like we did here?  You could
> > do it a bunch of different ways:
> > 
> > 1: if (ret < 0 || ret < ARRAY_SIZE(xxx))
> > 2: if (ret < (int)ARRAY_SIZE(xxx))
> > 3: if (ret != ARRAY_SIZE(xxx))
> > 
> > They're all equivalent.  I guess I don't like casting too much.  My first
> > approach to fixing this was just to declare if_version_length as an int, but
> > then I saw that originally there was a "bytes_recv < 0"
> > check and decided to go that way instead.
> 
> Actually bytes_recv should be probably of ssize_t type,  so could be the if_version_length. 
> 
> How did you find this, I haven't seen it in reported by sparse, smatch and I believe -Wsign-compare is suppressed in compilation warnings.

It's a new thing.  Julia noticed this kind of bug first and I have been
mucking around with it in Smatch as well.  My Smatch check has too many
false positives to publish right now because it thinks a some common
functions like ffs() return negative error codes.

regards,
dan carpenter