Re: [PATCH 1/4] media: sunxi-cir: Clean up dead register writes

[Sean Young]

On Tue, Jan 12, 2021 at 10:51:29PM -0600, Samuel Holland wrote:
> The register writes during driver removal occur after the device is
> already put back in reset, so they never had any effect.
> 
> Signed-off-by: Samuel Holland <samuel@xxxxxxxxxxxx>
> ---
>  drivers/media/rc/sunxi-cir.c | 10 ----------
>  1 file changed, 10 deletions(-)
> 
> diff --git a/drivers/media/rc/sunxi-cir.c b/drivers/media/rc/sunxi-cir.c
> index 8555c7798706..0a7f7eab3cc3 100644
> --- a/drivers/media/rc/sunxi-cir.c
> +++ b/drivers/media/rc/sunxi-cir.c
> @@ -342,22 +342,12 @@ static int sunxi_ir_probe(struct platform_device *pdev)
>  
>  static int sunxi_ir_remove(struct platform_device *pdev)
>  {
> -	unsigned long flags;
>  	struct sunxi_ir *ir = platform_get_drvdata(pdev);
>  
>  	clk_disable_unprepare(ir->clk);
>  	clk_disable_unprepare(ir->apb_clk);
>  	reset_control_assert(ir->rst);
>  
> -	spin_lock_irqsave(&ir->ir_lock, flags);
> -	/* disable IR IRQ */
> -	writel(0, ir->base + SUNXI_IR_RXINT_REG);
> -	/* clear All Rx Interrupt Status */
> -	writel(REG_RXSTA_CLEARALL, ir->base + SUNXI_IR_RXSTA_REG);
> -	/* disable IR */
> -	writel(0, ir->base + SUNXI_IR_CTL_REG);
> -	spin_unlock_irqrestore(&ir->ir_lock, flags);
> -
>  	rc_unregister_device(ir->rc);
>  	return 0;
>  }

I don't think there is anything wrong with the patch, however here the
driver does rc_unregister_device() *after* disabling it. Userspace can
still hold a file descriptor open, and call e.g. LIRC_SET_REC_TIMEOUT
ioctl, which causes various writes the sunxi-cir registers.

The order should be reversed.


Sean