mn10300: READ_ONCE() now implies smp_read_barrier_depends()

Given that READ_ONCE() now implies smp_read_barrier_depends(),
there is no need for the open-coded smp_read_barrier_depends() in
mn10300_serial_receive_interrupt() and mn10300_serial_poll_get_char().
This commit therefore removes them, but replaces them with comments
calling out that carrying dependencies through non-pointers is quite
dangerous.  Compilers simply know too much about integers.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: <linux-am33-list@redhat.com>

diff --git a/arch/mn10300/kernel/mn10300-serial.c b/arch/mn10300/kernel/mn10300-serial.c
index d7ef123..4994b57 100644 (file)
--- a/arch/mn10300/kernel/mn10300-serial.c
+++ b/arch/mn10300/kernel/mn10300-serial.c
@@ -550,7 +550,7 @@ try_again:
                return;
        }
 
-       smp_read_barrier_depends();
+       /* READ_ONCE() enforces dependency, but dangerous through integer!!! */
        ch = port->rx_buffer[ix++];
        st = port->rx_buffer[ix++];
        smp_mb();
@@ -1728,7 +1728,10 @@ static int mn10300_serial_poll_get_char(struct uart_port *_port)
                        if (CIRC_CNT(port->rx_inp, ix, MNSC_BUFFER_SIZE) == 0)
                                return NO_POLL_CHAR;
 
-                       smp_read_barrier_depends();
+                       /*
+                        * READ_ONCE() enforces dependency, but dangerous
+                        * through integer!!!
+                        */
                        ch = port->rx_buffer[ix++];
                        st = port->rx_buffer[ix++];
                        smp_mb();