Re: [PATCH] drm/mediatek: fix a loop timeout

[Philipp Zabel <p.zabel@pengutronix.de>]

Hi Dan,

On Thu, 2017-04-20 at 13:24 +0300, Dan Carpenter wrote:
> We expected to end with "timeout_ms" set to zero, but it's a post-op so
> the current code ends with -1.  Let's fix this by changing it to a
> pre-op.
> 
> Fixes: 21898816831f ("drm/mediatek: add dsi transfer function")
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>

Indeed without this patch the timeout error path is only ever chosen if
the busy bit happened to flip exactly in the last iteration of the loop.

Acked-by: Philipp Zabel <p.zabel@pengutronix.de>

> diff --git a/drivers/gpu/drm/mediatek/mtk_dsi.c b/drivers/gpu/drm/mediatek/mtk_dsi.c
> index 808b995a990f..3a0b6d1057a2 100644
> --- a/drivers/gpu/drm/mediatek/mtk_dsi.c
> +++ b/drivers/gpu/drm/mediatek/mtk_dsi.c
> @@ -902,7 +902,7 @@ static void mtk_dsi_wait_for_idle(struct mtk_dsi *dsi)
>  {
>  	u32 timeout_ms = 500000; /* total 1s ~ 2s timeout */
>  
> -	while (timeout_ms--) {
> +	while (--timeout_ms) {
>  		if (!(readl(dsi->regs + DSI_INTSTA) & DSI_BUSY))
>  			break;
>  
> 

Instead of just calling usleep_range(2, 4) up to 500000 times in this
open coded loop, this should be replaced with a single call to
readl_poll_timeout() though.

regards
Philipp