Re: include/net/sock.h:2100:16: sparse: sparse: cast to non-scalar

[Al Viro <viro@zeniv.linux.org.uk>]

On Sat, Jan 10, 2026 at 10:15:09PM +0000, Al Viro wrote:
> On Sun, Jan 11, 2026 at 05:06:31AM +0800, kernel test robot wrote:
> 
> Folks, could your bot be taught to recognize that kind of situation?
> It's very clear what's going on - an inline function that gets the
> same warning with this config every time it is called.  Commit in
> question has added several calls of that thing; the origin of warning
> has nothing whatsoever to do with it, and that can be guessed without
> even looking at the source - the quoted diff between the logs is quite
> enough for that.
> 
> Looking at the function in question,

[accidentally sent halfway through editing, sorry]

e84a4927a404f (Eric Dumazet               2025-06-20 13:30:00 +0000 2097) static inline kuid_t sk_uid(const struct sock *sk)
e84a4927a404f (Eric Dumazet               2025-06-20 13:30:00 +0000 2098) {
e84a4927a404f (Eric Dumazet               2025-06-20 13:30:00 +0000 2099)       /* Paired with WRITE_ONCE() in sockfs_setattr() */
e84a4927a404f (Eric Dumazet               2025-06-20 13:30:00 +0000 2100)       return READ_ONCE(sk->sk_uid);
e84a4927a404f (Eric Dumazet               2025-06-20 13:30:00 +0000 2101) }

we see that warning in question almost certainly comes from
	READ_ONCE(some_kuid_t_object)
and looking at alpha we see this:
#define __READ_ONCE(x)                                                  \
({                                                                      \
        __unqual_scalar_typeof(x) __x =                                 \
		(*(volatile typeof(__x) *)(&(x)));                      \
	mb();                                                           \
	(typeof(x))__x;                                                 \
})
combined with
typedef struct {
        uid_t val;
} kuid_t;

IOW, it complains about a cast from structure to itself, which is fair
enough - C is pretty clear about not allowing any typecasts to or from
non-scalar types, tautological or not.

Why do we even need that cast?  Seeing that generic __READ_ONCE() is
#define __READ_ONCE(x)  (*(const volatile __unqual_scalar_typeof(x) *)&(x))
the cast added on alpha seems to be pointless.

Folks involved in putting that cast in arch/alpha/include/asm/rwonce.h Cc'd...