Re: [rfc] increase struct page size?!

[Eric Dumazet]

On Mon, 21 May 2007 01:08:13 -0700
William Lee Irwin III <wli@xxxxxxxxxxxxxx> wrote:

> Now that I've been informed of the ->_count and ->_mapcount issues,
> I'd say that they're grave and should be corrected even at the cost
> of sizeof(struct page).


As long we handle 4 KB pages, adding 64 bits per page means 0.2 % of overhead. Ouch...

We currently have an overhead of 1.36 % for mem_map

Maybe we can still use 32 bits counters, and make sure non root users cannot
make these counters exceed 2^30. (I believe high order bit has already a meaning, 
check page_mapped() definition)

We could use a special atomic_inc_if_not_huge() function, that could revert to
 normal atomic_inc() on machines with less than 32 GB (using alternative_() variant)

On small setups (or 32 bits arches), atomic_inc_if_not_huge() would unconditionnally 
increment the counter.

#if !defined(BIG_MACHINES)
static int inline atomic_inc_if_not_huge(atomic_t *v)
{
atomic_inc(v);
return 1;
}
#else
extern int atomic_inc_if_not_huge(atomic_t *v);

#endif


/* in a .c file */
/* could be patched at boot time if available memory < 32GB (or other limit) */
#if defined(BIG_MACHINES)
#define MAP_LIMIT_COUNT (2<<30)
int atomic_inc_if_not_huge(atomic_t *v);
{
/* lazy test, we dont care enough to do a real atomic read-modify-write */
if (unlikely(atomic_read(v) >= MAP_LIMIT_COUNT)) {
	if (non_root_user())
		return 0;
	}
atomic_inc(v);
return 1;
}
#endif
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/