Re: [PATCH] platform/x86: fujitsu-laptop: Revert UNSUPPORTED_CMD back to an int

[Jonathan Woithe]

On Sat, Mar 10, 2018 at 09:43:53PM +0100, Micha?? K??pie?? wrote:
> UNSUPPORTED_CMD was previously 0x80000000 (int), but commit 819cddae7cfa
> ("platform/x86: fujitsu-laptop: Clean up constants") changed it into an
> unsigned long due to BIT() being used to define it.  As call_fext_func()
> returns an int, 0x80000000 would get type promoted when compared to an
> unsigned long, which on a 64-bit system would cause it to become
> 0xffffffff80000000 due to sign extension.  This causes one logical
> condition in fujitsu-laptop to always be true and another one to always
> be false on 64-bit systems.  Fix this by reverting UNSUPPORTED_CMD back
> to an int.
> 
> This patch fixes the following smatch warnings:
> 
> drivers/platform/x86/fujitsu-laptop.c:763 acpi_fujitsu_laptop_leds_register() warn: always true condition '(call_fext_func(device, ((1 << (12)) | (1 << (0))), 2, (1 << (16)), 0) != (1 << (31))) => (s32min-s32max != 2147483648)'
> drivers/platform/x86/fujitsu-laptop.c:816 acpi_fujitsu_laptop_add() warn: impossible condition '(priv->flags_supported == (1 << (31))) => (0-2147483647,18446744071562067968-u64max == 2147483648)'
> 
> Fixes: 819cddae7cfa ("platform/x86: fujitsu-laptop: Clean up constants")
> Reported-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> Signed-off-by: Micha?? K??pie?? <kernel@xxxxxxxxxx>
> ---
> This fixes a bug introduced by a commit queued for 4.17, so it needs to
> be applied on top of for-next.
> 
>  drivers/platform/x86/fujitsu-laptop.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/platform/x86/fujitsu-laptop.c b/drivers/platform/x86/fujitsu-laptop.c
> index 13bcdfea5349..6f4a55a53ced 100644
> --- a/drivers/platform/x86/fujitsu-laptop.c
> +++ b/drivers/platform/x86/fujitsu-laptop.c
> @@ -85,7 +85,7 @@
>  #define FUNC_BACKLIGHT			(BIT(12) | BIT(2))
>  
>  /* FUNC interface - responses */
> -#define UNSUPPORTED_CMD			BIT(31)
> +#define UNSUPPORTED_CMD			0x80000000
>  
>  /* FUNC interface - status flags */
>  #define FLAG_RFKILL			BIT(5)

This looks like a sensible, succinct solution to the regression.

Reviewed-by: Jonathan Woithe <jwoithe@xxxxxxxxxx>

Regards
  jonathan