Re: [PATCH net-next v3 1/4] net: dsa: yt921x: Use *_ULL bitfield macros for VLAN_CTRL

[Russell King (Oracle)]

On Fri, Nov 28, 2025 at 10:51:41AM +0000, david laight wrote:
> On Wed, 26 Nov 2025 17:32:34 +0800
> David Yang <mmyangfl@xxxxxxxxx> wrote:
> 
> > VLAN_CTRL should be treated as a 64-bit register. GENMASK and BIT
> > macros use unsigned long as the underlying type, which will result in a
> > build error on architectures where sizeof(long) == 4.
> 
> I suspect GENMASK() should generate u32 or u64 depending on the value
> of a constant 'high bit'.

I suggest checking before making such statements to save embarrasment.
The above is incorrect.

#define GENMASK_TYPE(t, h, l)                                   \
        ((t)(GENMASK_INPUT_CHECK(h, l) +                        \
             (type_max(t) << (l) &                              \
              type_max(t) >> (BITS_PER_TYPE(t) - 1 - (h)))))

#define GENMASK(h, l)           GENMASK_TYPE(unsigned long, h, l)
#define GENMASK_ULL(h, l)       GENMASK_TYPE(unsigned long long, h, l)

#define GENMASK_U8(h, l)        GENMASK_TYPE(u8, h, l)
#define GENMASK_U16(h, l)       GENMASK_TYPE(u16, h, l)
#define GENMASK_U32(h, l)       GENMASK_TYPE(u32, h, l)
#define GENMASK_U64(h, l)       GENMASK_TYPE(u64, h, l)
#define GENMASK_U128(h, l)      GENMASK_TYPE(u128, h, l)

Note that GENMASK(33, 15) will fail to build on 32-bit systems.

> I found code elsewhere that doesn't really want FIELD_PREP() to
> generate a 64bit value.
> 
> There are actually a lot of dubious uses of 'long' throughout
> the kernel that break on 32bit.
> (Actually pretty much all of them!)

If you're referring to the use of GENMASK() with bitfields larger
than 32-bits, then as can be seen from the above, the code wouldn't
even compile and our CI systems would be screaming about it. They
aren't, so I think your statement here is also wrong.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!