lkml.org 
[lkml]   [2018]   [Dec]   [31]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
From
Date
SubjectRe: [PATCH v8 20/25] powerpc, fbdev: Use arch_nvram_ops methods instead of nvram_read_byte() and nvram_write_byte()
On Sun, Dec 30, 2018 at 12:43 AM Finn Thain <fthain@telegraphics.com.au> wrote:

>
> Is there some benefit, or is that just personal taste?
>
> Avoiding changes to call sites avoids code review, but I think 1) the
> thinkpad_acpi changes have already been reviewed and 2) the fbdev changes
> need review anyway.
>
> Your suggesion would add several new entities and one extra layer of
> indirection.
>
> I think indirection harms readability because now the reader now has to go
> and look up the meaning of the new entities.
>
> It's not the case that we need to choose between definitions of
> nvram_read_byte() at compile time, or stub them out:
>
> #ifdef CONFIG_FOO
> static inline unsigned char nvram_read_byte(int addr)
> {
> return arch_nvram_ops.read_byte(addr);
> }
> #else
> static inline unsigned char nvram_read_byte(int addr) { }
> #endif
>
> And I don't anticipate a need for a macro here either:
>
> #define nvram_read_byte(a) random_nvram_read_byte_impl(a)
>
> I think I've used the simplest solution.

Having the indirection would help if the inline function can
encapsulate the NULL pointer check, like

static inline unsigned char nvram_read_byte(loff_t addr)
{
char data;

if (!IS_ENABLED(CONFIG_NVRAM))
return 0xff;

if (arch_nvram_ops.read_byte)
return arch_nvram_ops.read_byte(addr);

if (arch_nvram_ops.read)
return arch_nvram_ops.read(char, 1, &addr);

return 0xff;
}

(the above assumes no #ifdef in the structure definition, if you
keep the #ifdef there they have to be added here as well).

Arnd

\
 
 \ /
  Last update: 2018-12-31 13:30    [W:0.095 / U:0.444 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site