Re: [PATCH] Re-organization of PIDMAP_ENTRIES macro expansion

[Randy Dunlap <randy.dunlap@oracle.com>]

On Mon, 14 Jan 2008 16:20:35 +0530 Ratnadeep Joshi wrote:

> This patch tries to re-organize the macro expansion of PIDMAP_ENTRIES
> (possibly) to a more clear one.
> 
> Thanks,
> - Ratnadeep Joshi
> 
> diff --git a/include/linux/pid_namespace.h
> b/include/linux/pid_namespace.h
> index 1689e28..06e3e99 100644
> --- a/include/linux/pid_namespace.h
> +++ b/include/linux/pid_namespace.h
> @@ -12,7 +12,7 @@ struct pidmap {
>         void *page;
>  };
>  
> -#define PIDMAP_ENTRIES         ((PID_MAX_LIMIT + 8*PAGE_SIZE -
> 1)/PAGE_SIZE/8)
> +#define PIDMAP_ENTRIES         ((PID_MAX_LIMIT - 1)/PAGE_SIZE/8 + 1)
>  
>  struct pid_namespace {
>  	struct kref kref;


I beg to disagree.  There is a very common (idiomatic) way for
doing this kind of calculation and the first/original expression
is in that form, although it's a little difficult to recognize.

This common formula lives in linux/kernel.h:

#define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))

where n is (AFAIK) numerator and d is divisor.

Example:  One use of this is in linux/bitops.h:

#define BITS_TO_LONGS(nr)	DIV_ROUND_UP(nr, BITS_PER_LONG)

This macro returns the number of longs that it takes to hold
'nr' bits.  Expanded, it looks like:

nr_longs = (nr_bits + BITS_PER_LONG - 1) / BITS_PER_LONG;

The idiomatic formula handles boundary conditions very nicely.


Now for PIDMAP_ENTRIES, use 8 * PAGE_SIZE as the (d) part of
DIV_ROUND_UP() and you can see that it is done correctly.

---
~Randy