On Tue, Oct 16, 2018 at 19:10:03 +0800, guangrong.x...@gmail.com wrote:
(snip)
> diff --git a/include/qemu/ptr_ring.h b/include/qemu/ptr_ring.h
> new file mode 100644
> index 0000000000..d8266d45f6
> --- /dev/null
> +++ b/include/qemu/ptr_ring.h
> @@ -0,0 +1,235 @@
(snip)
> +#define SMP_CACHE_BYTES 64
> +#define ____cacheline_aligned_in_smp \
> + __attribute__((__aligned__(SMP_CACHE_BYTES)))
You could use QEMU_ALIGNED() here.
> +
> +#define WRITE_ONCE(ptr, val) \
> + (*((volatile typeof(ptr) *)(&(ptr))) = (val))
> +#define READ_ONCE(ptr) (*((volatile typeof(ptr) *)(&(ptr))))
Why not atomic_read/set, like in the rest of the QEMU code base?
Furthermore, READ_ONCE in the kernel implies smp_read_barrier_depends,
whereas here you're not bringing that in. That means that your
barrier pairing, e.g.
> + /* Pairs with READ_ONCE in ptr_ring_consume. */
> + smp_wmb();
is incorrect for Alpha.
Thanks,
E.
