Hi Sergei,
On Thu, Aug 30, 2018 at 9:19 PM Sergei Shtylyov
<[email protected]> wrote:
> When trying to use CMT for clockevents on R-Car gen3 SoCs, I noticed that
> the maximum delta (in ns) for the broadcast timer was diplayed as 1000 in
> /proc/timer_list. It turned out that when calculating it, the driver did
> shift left 1 by 32 (causing what I think was undefined behaviour) getting
> 1 as a result. Using 1UL instead fixed the maximum delta and did even fix
> switching an active clocksource to a CMT channel (which caused the system
> to go non-interactive before).
>
> Signed-off-by: Sergei Shtylyov <[email protected]>
>
> ---
> This patch is against the 'tip.git' repo's 'timers/core' branch.
>
> I am not sure whether the CMT driver was ever used on SH64; on R-Car gen3
> (ARM64) I'm only enabling this driver now, so not sure how urgent is this...
>
> drivers/clocksource/sh_cmt.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> Index: tip/drivers/clocksource/sh_cmt.c
> ===================================================================
> --- tip.orig/drivers/clocksource/sh_cmt.c
> +++ tip/drivers/clocksource/sh_cmt.c
> @@ -891,7 +891,7 @@ static int sh_cmt_setup_channel(struct s
> if (cmt->info->width == (sizeof(ch->max_match_value) * 8))
> ch->max_match_value = ~0;
> else
> - ch->max_match_value = (1 << cmt->info->width) - 1;
> + ch->max_match_value = (1UL << cmt->info->width) - 1;
>
> ch->match_value = ch->max_match_value;
> raw_spin_lock_init(&ch->lock);
Shouldn't all those fields/variables be changed from "unsigned long" to "u32",
as they really should match the CMT's register sizes?
Gr{oetje,eeting}s,
Geert
--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]
In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds
