Re: [PATCH 03/18] arm64/uaccess: fix sparse errors

[Michael S. Tsirkin]

On Mon, Dec 15, 2014 at 11:23:11AM +0000, Russell King - ARM Linux wrote:
> On Mon, Dec 15, 2014 at 11:17:16AM +0000, Will Deacon wrote:
> > On Sun, Dec 14, 2014 at 04:52:09PM +0000, Michael S. Tsirkin wrote:
> > > virtio wants to read bitwise types from userspace using get_user.  At the
> > > moment this triggers sparse errors, since the value is passed through an
> > > integer.
> > > 
> > > Fix that up using __force.
> > > 
> > > Signed-off-by: Michael S. Tsirkin <mst@xxxxxxxxxx>
> > > ---
> > >  arch/arm64/include/asm/uaccess.h | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/arch/arm64/include/asm/uaccess.h b/arch/arm64/include/asm/uaccess.h
> > > index 3bf8f4e..8d66bcf 100644
> > > --- a/arch/arm64/include/asm/uaccess.h
> > > +++ b/arch/arm64/include/asm/uaccess.h
> > > @@ -147,7 +147,7 @@ do {									\
> > >  	default:							\
> > >  		BUILD_BUG();						\
> > >  	}								\
> > > -	(x) = (__typeof__(*(ptr)))__gu_val;				\
> > > +	(x) = (__force __typeof__(*(ptr)))__gu_val;				\
> > >  } while (0)
> > >  
> > >  #define __get_user(x, ptr)						\
> > 
> > Acked-by: Will Deacon <will.deacon@xxxxxxx>
> 
> This also means you can do stuff like:
> 
> 	u32 *p;
> 	__le32 v;
> 
> 	err = get_user(p, v);
> 
> which is not right.  Both the dereferenced pointer type and the destination
> type should be compatible, and if one is a bitwise type but the other isn't,
> that seems like a valid case to warn.

I just verified this case:
#define __force __attribute__((force))
#define __bitwise__ __attribute__((bitwise))
#define get_user(x, ptr) \
do {\
        unsigned long __gu_val = 0; \
       (x) = (__force __typeof__(*(ptr)))__gu_val;  \
} while (0)

typedef unsigned u32;
typedef u32 __bitwise__ __le32;

static u32 *p;
static __le32 v;


int main(int argc, char **argv)
{
        get_user(v, p);
        return 0;
}


Produces a warning as expected.

So I think the above comment is a result of a mistake.
Can you confirm please?



> I don't see any use of get_user() in drivers/virtio in mainline, so I can't
> check further.


this is an example of the case which I'm fixing:

#define __force __attribute__((force))
#define __bitwise__ __attribute__((bitwise))
#define get_user(x, ptr) \
do {\
        unsigned long __gu_val = 0; \
       (x) = (__typeof__(*(ptr)))__gu_val;  \
} while (0)

typedef unsigned u32;
typedef u32 __bitwise__ __le32;

static __le32 *p;
static __le32 v;


int main(int argc, char **argv)
{
        get_user(v, p);
        return 0;
}

the code is correct but produces a warning:
a.c:18:9: warning: cast to restricted __le32


The cast near __typeof__ above needs __force,
this is what my patch does.



> -- 
> FTTC broadband for 0.8mile line: currently at 9.5Mbps down 400kbps up
> according to speedtest.net.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/