how to purposfully fragment memory?

how to purposfully fragment memory?

[Chris Friesen]

I'm trying to test code paths dealing with fragmented memory, so I'd 
like to have a simple way to cause fragmented memory in the kernel.  Is 
there any API in the kernel that would let me allocate two contiguous 
pages, then free one of them?


I tried the following, but it triggers an oops in free_page():

unsigned long addr =
	__get_free_pages(GFP_KERNEL|__GFP_NORETRY|__GFP_NOWARN, 1);
if (!addr)
	break;
free_page(addr+4096);


Thanks,

Chris

Re: how to purposfully fragment memory?

[Dan Kruchinin]

On 3/20/07, Chris Friesen <cfriesen@nortel.com> wrote:
> Asgard wrote:
>
> > btw, can you describe your aim in details?
>
> I'm doing some work on a network device driver.  We were having jumbo
> frame issues after being up for a month or so because it couldn't
> allocate enough contiguous memory.
>
> We've got a new code path now that doesn't need large contiguous
> allocations, so I'd like to demonstrate it by forcing the system to
> become severely fragmented.
>
> Chris
>

and by the way, you can try to use void split_page(struct page *page,
unsigned int order) function, but keep in mind, that it's not
recommended to use it in drivers:
/*
 * split_page takes a non-compound higher-order page, and splits it into
 * n (1<<order) sub-pages: page[0..n]
 * Each sub-page must be freed individually.
 *
 * Note: this is probably too low level an operation for use in drivers.
 * Please consult with lkml before using this in your driver.
 */
void split_page(struct page *page, unsigned int order)

Good luck.
Dan.