On Fri, Mar 13, 2020 at 07:54:49AM -0700, Bart Van Assche wrote: > On 2020-03-13 02:15, Andy Shevchenko wrote: > > On Thu, Mar 12, 2020 at 07:37:16PM -0700, Bart Van Assche wrote: > >> +static inline void __put_unaligned_be24(u32 val, u8 *p) > > > > const u32 val > > Hi Andy, > > Thanks for the review. The above suggestion surprises me: as far as I > can tell almost nobody declares function arguments that are passed by > value as 'const' in the Linux kernel: > > $ git grep -nH '(const[^\*,]*,' | wc -l > 1065 > > That number is negligible compared to the number of function declarations: > > $ git grep -nH '(.*);$' | wc -l > 2692721 It's a surprising "argument". If 100500 do worse it doesn't mean 3000 shouldn't do it better. And of course first grep is incomplete and second one too broad. Just for (generic) headers: $ git grep -n '[a-z_0-9]([^)]*\bconst [^)]\+)' -- include | wc -l 4342 $ git grep -n '[a-z_0-9]([^)]\+)' -- include | wc -l 69672 ~6% in headers. I don't think it's negligible. You have at least two advantages on this: a) we really don't modify the content of the input value; b) it will be consistent with the rest of consolidated helpers. -- With Best Regards, Andy Shevchenko
