On Fri, Sep 8, 2023, at 08:51, Geert Uytterhoeven wrote:
On Fri, Sep 8, 2023 at 1:37 AM Finn Thain <fthain@xxxxxxxxxxxxxx> wrote:
On Thu, 7 Sep 2023, Geert Uytterhoeven wrote:
--- a/arch/m68k/apollo/config.c
+++ b/arch/m68k/apollo/config.c
@@ -146,13 +146,11 @@ void __init config_apollo(void)
irqreturn_t dn_timer_int(int irq, void *dev_id)
{
- volatile unsigned char x;
-
legacy_timer_tick(1);
timer_heartbeat();
- x = *(volatile unsigned char *)(apollo_timer + 3);
- x = *(volatile unsigned char *)(apollo_timer + 5);
+ READ_ONCE(*(volatile unsigned char *)(apollo_timer + 3));
+ READ_ONCE(*(volatile unsigned char *)(apollo_timer + 5));
return IRQ_HANDLED;
}
I believe the volatile cast is redundant here, as READ_ONCE does that.
Yes it does. I didn't drop the volatile out of fear of introducing
some higher W-level warning about casting away volatility, but upon
closer look, the apollo_timer definition itself does not use volatile.
I think ideally you'd want apollo_timer or IO_BASE to be marked __iomem
and passed to readb() instead. OTOH, I don't think anyone actually
cares about apollo, so your version is probably good enough.
Arnd
