On Wed, Mar 15, 2017 at 01:31:51PM +0000, Jon Hunter wrote:
> If we fail to map the IO space for the Tegra flowctrl module in
> tegra_flowctrl_init(), then if any of the public APIs for the Tegra
> flowctrl driver are called we still still attempt to access the
> hardware. Fix this by only accessing the hardware if the
> 'tegra_flowctrl_base' address has not been setup correctly and warn
> if we attempt to access the hardware when it has not.
>
> Signed-off-by: Jon Hunter <jonathanh@xxxxxxxxxx>
> ---
> arch/arm/mach-tegra/flowctrl.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/arch/arm/mach-tegra/flowctrl.c b/arch/arm/mach-tegra/flowctrl.c
> index 475e783992fd..40b15e39f849 100644
> --- a/arch/arm/mach-tegra/flowctrl.c
> +++ b/arch/arm/mach-tegra/flowctrl.c
> @@ -47,6 +47,9 @@ static void __iomem *tegra_flowctrl_base;
>
> static void flowctrl_update(u8 offset, u32 value)
> {
> + if (WARN_ONCE(!tegra_flowctrl_base, "Tegra flowctrl not supported!"))
Technically I think this string needs a terminating "\n" because it is
eventually passed to printk(), even though the WARN_ONCE() will probably
format the output correctly even without it.
Also, "not supported" isn't a very accurate description. This has to do
with the flow controller not having been initialized yet.
Thierry
Attachment:
signature.asc
Description: PGP signature
