Re: [PATCH 1/2] serial: 8250: Let drivers request full 16550A feature probing

[Jiri Slaby <jirislaby@kernel.org>]

On 19. 09. 22, 10:18, Maciej W. Rozycki wrote:
> On Mon, 19 Sep 2022, Jiri Slaby wrote:
> 
>>> --- linux-macro.orig/include/linux/serial_core.h
>>> +++ linux-macro/include/linux/serial_core.h
>>> @@ -414,7 +414,7 @@ struct uart_icount {
>>>    	__u32	buf_overrun;
>>>    };
>>>    -typedef unsigned int __bitwise upf_t;
>>> +typedef __u64 __bitwise upf_t;
>>
>> Why __u64 and not u64?
> 
>   For consistency as there's `__u32' used elsewhere in this file.  It's not
> clear to me what our rules WRT the use of `s*'/`u*' vs `__s*'/`__u*' are.
> I don't think we have it mentioned under Documentation/.  Please clarify
> if you know and I can update the change accordingly.

The rule is, AFAICT, use __u*/__s* in user (uapi) headers. Everywhere 
else, use u*/s*.

>>> @@ -522,6 +522,7 @@ struct uart_port {
>>>    #define UPF_FIXED_PORT		((__force upf_t) (1 << 29))
>>>    #define UPF_DEAD		((__force upf_t) (1 << 30))
>>>    #define UPF_IOREMAP		((__force upf_t) (1 << 31))
>>> +#define UPF_FULL_PROBE		((__force upf_t) (1ULL << 32))
>>
>> This looks like a perfect time to switch them all to BIT_ULL().
> 
>   Good point, I keep forgetting about that macro.  I'll keep this part
> unchanged for the purpose of backporting and add 3/3 to switch all the
> macros over.

OK.

thanks,
-- 
js
suse labs