On Tue, Mar 15, 2022 at 09:15:34AM +0000, Visa Hankala wrote:
> However, some DEC Alpha CPUs have their data caches divided into cache
> banks to improve bandwidth. These cache banks are relatively
> independent. The system maintains coherency, but bus contention can
> delay propagation of cache updates. If the loads spanned different cache
> banks, the second load could deliver data which is older than the
> initial load's value. The data dependency barrier causes an interlock
> with cache updating, ensuring causal ordering.)
The code with the membar is copied from READ_ONCE() which is copied
from Linux. The membar_datadep_consumer() has an #ifdef __alpha__
in it. It is only used for that case. I don't know whether we
want to support such CPU. But if that is the case, we need the
membar.
What do you need refcnt_read() for? Is it only for assert? Then
a refcnt_assert() without membar or atomic_load might be better.
bluhm
> Index: sys/sys/atomic.h
> ===================================================================
> RCS file: src/sys/sys/atomic.h,v
> retrieving revision 1.8
> diff -u -p -r1.8 atomic.h
> --- sys/sys/atomic.h 11 Mar 2022 19:02:15 -0000 1.8
> +++ sys/sys/atomic.h 15 Mar 2022 07:52:39 -0000
> @@ -201,26 +201,16 @@ atomic_sub_long_nv(volatile unsigned lon
> * atomic_load_* - read from memory
> */
>
> -static inline void membar_datadep_consumer(void);
> -
> static inline unsigned int
> atomic_load_int(volatile unsigned int *p)
> {
> - unsigned int v;
> -
> - v = *p;
> - membar_datadep_consumer();
> - return v;
> + return *p;
> }
>
> static inline unsigned long
> atomic_load_long(volatile unsigned long *p)
> {
> - unsigned long v;
> -
> - v = *p;
> - membar_datadep_consumer();
> - return v;
> + return *p;
> }
>
> /*
