Re: [RFC][PATCH 2/3] a help function for find physically contiguous block.

[Bob Liu]

On Tue, Oct 26, 2010 at 6:04 PM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote:
> From: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
>
> Unlike memory hotplug, at an allocation of contigous memory range, address
> may not be a problem. IOW, if a requester of memory wants to allocate 100M of
> of contigous memory, placement of allocated memory may not be a problem.
> So, "finding a range of memory which seems to be MOVABLE" is required.
>
> This patch adds a functon to isolate a length of memory within [start, end).
> This function returns a pfn which is 1st page of isolated contigous chunk
> of given length within [start, end).
>
> After isolation, free memory within this area will never be allocated.
> But some pages will remain as "Used/LRU" pages. They should be dropped by
> page reclaim or migration.
>
> Changelog:
> Â- zone is added to the argument.
> Â- fixed a case that zones are not in linear.
> Â- added zone->lock.
>
>
> Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx>
> ---
> Âmm/page_isolation.c | Â148 ++++++++++++++++++++++++++++++++++++++++++++++++++++
> Â1 file changed, 148 insertions(+)
>
> Index: mmotm-1024/mm/page_isolation.c
> ===================================================================
> --- mmotm-1024.orig/mm/page_isolation.c
> +++ mmotm-1024/mm/page_isolation.c
> @@ -7,6 +7,7 @@
> Â#include <linux/pageblock-flags.h>
> Â#include <linux/memcontrol.h>
> Â#include <linux/migrate.h>
> +#include <linux/memory_hotplug.h>
> Â#include <linux/mm_inline.h>
> Â#include "internal.h"
>
> @@ -250,3 +251,150 @@ int do_migrate_range(unsigned long start
> Âout:
> Â Â Â Âreturn ret;
> Â}
> +
> +/*
> + * Functions for getting contiguous MOVABLE pages in a zone.
> + */
> +struct page_range {
> + Â Â Â unsigned long base; /* Base address of searching contigouous block */
> + Â Â Â unsigned long end;
> + Â Â Â unsigned long pages;/* Length of contiguous block */
> + Â Â Â int align_order;
> + Â Â Â unsigned long align_mask;
> +};
> +
> +int __get_contig_block(unsigned long pfn, unsigned long nr_pages, void *arg)
> +{
> + Â Â Â struct page_range *blockinfo = arg;
> + Â Â Â unsigned long end;
> +
> + Â Â Â end = pfn + nr_pages;
> + Â Â Â pfn = ALIGN(pfn, 1 << blockinfo->align_order);
> + Â Â Â end = end & ~(MAX_ORDER_NR_PAGES - 1);
> +
> + Â Â Â if (end < pfn)
> + Â Â Â Â Â Â Â return 0;
> + Â Â Â if (end - pfn >= blockinfo->pages) {
> + Â Â Â Â Â Â Â blockinfo->base = pfn;
> + Â Â Â Â Â Â Â blockinfo->end = end;
> + Â Â Â Â Â Â Â return 1;
> + Â Â Â }
> + Â Â Â return 0;
> +}
> +
> +static void __trim_zone(struct zone *zone, struct page_range *range)
> +{
> + Â Â Â unsigned long pfn;
> + Â Â Â /*
> + Â Â Â Â* skip pages which dones'nt under the zone.
> + Â Â Â Â* There are some archs which zones are not in linear layout.
> + Â Â Â Â*/
> + Â Â Â if (page_zone(pfn_to_page(range->base)) != zone) {
> + Â Â Â Â Â Â Â for (pfn = range->base;
> + Â Â Â Â Â Â Â Â Â Â Â pfn < range->end;
> + Â Â Â Â Â Â Â Â Â Â Â pfn += MAX_ORDER_NR_PAGES) {
> + Â Â Â Â Â Â Â Â Â Â Â if (page_zone(pfn_to_page(pfn)) == zone)
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
> + Â Â Â Â Â Â Â }
> + Â Â Â Â Â Â Â range->base = min(pfn, range->end);
> + Â Â Â }
> + Â Â Â /* Here, range-> base is in the zone if range->base != range->end */
> + Â Â Â for (pfn = range->base;
> + Â Â Â Â Â Âpfn < range->end;
> + Â Â Â Â Â Âpfn += MAX_ORDER_NR_PAGES) {
> + Â Â Â Â Â Â Â if (zone != page_zone(pfn_to_page(pfn))) {
> + Â Â Â Â Â Â Â Â Â Â Â pfn = pfn - MAX_ORDER_NR_PAGES;
> + Â Â Â Â Â Â Â Â Â Â Â break;
> + Â Â Â Â Â Â Â }
> + Â Â Â }
> + Â Â Â range->end = min(pfn, range->end);
> + Â Â Â return;
> +}
> +
> +/*
> + * This function is for finding a contiguous memory block which has length
> + * of pages and MOVABLE. If it finds, make the range of pages as ISOLATED
> + * and return the first page's pfn.
> + * This checks all pages in the returned range is free of Pg_LRU. To reduce
> + * the risk of false-positive testing, lru_add_drain_all() should be called
> + * before this function to reduce pages on pagevec for zones.
> + */
> +
> +static unsigned long find_contig_block(unsigned long base,
> + Â Â Â Â Â Â Â unsigned long end, unsigned long pages,
> + Â Â Â Â Â Â Â int align_order, struct zone *zone)
> +{
> + Â Â Â unsigned long pfn, pos;
> + Â Â Â struct page_range blockinfo;
> + Â Â Â int ret;
> +
> + Â Â Â VM_BUG_ON(pages & (MAX_ORDER_NR_PAGES - 1));
> + Â Â Â VM_BUG_ON(base & ((1 << align_order) - 1));
> +retry:
> + Â Â Â blockinfo.base = base;
> + Â Â Â blockinfo.end = end;
> + Â Â Â blockinfo.pages = pages;
> + Â Â Â blockinfo.align_order = align_order;
> + Â Â Â blockinfo.align_mask = (1 << align_order) - 1;
> + Â Â Â /*
> + Â Â Â Â* At first, check physical page layout and skip memory holes.
> + Â Â Â Â*/
> + Â Â Â ret = walk_system_ram_range(base, end - base, &blockinfo,
> + Â Â Â Â Â Â Â __get_contig_block);
> + Â Â Â if (!ret)
> + Â Â Â Â Â Â Â return 0;
> + Â Â Â /* check contiguous pages in a zone */
> + Â Â Â __trim_zone(zone, &blockinfo);
> +
> + Â Â Â /*
> + Â Â Â Â* Ok, we found contiguous memory chunk of size. Isolate it.
> + Â Â Â Â* We just search MAX_ORDER aligned range.
> + Â Â Â Â*/
> + Â Â Â for (pfn = blockinfo.base; pfn + pages <= blockinfo.end;
> + Â Â Â Â Â Âpfn += (1 << align_order)) {
> + Â Â Â Â Â Â Â struct zone *z = page_zone(pfn_to_page(pfn));
> +
> + Â Â Â Â Â Â Â spin_lock_irq(&z->lock);
> + Â Â Â Â Â Â Â pos = pfn;
> + Â Â Â Â Â Â Â /*
> + Â Â Â Â Â Â Â Â* Check the range only contains free pages or LRU pages.
> + Â Â Â Â Â Â Â Â*/
> + Â Â Â Â Â Â Â while (pos < pfn + pages) {
> + Â Â Â Â Â Â Â Â Â Â Â struct page *p;
> +
> + Â Â Â Â Â Â Â Â Â Â Â if (!pfn_valid_within(pos))
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
> + Â Â Â Â Â Â Â Â Â Â Â p = pfn_to_page(pos);
> + Â Â Â Â Â Â Â Â Â Â Â if (PageReserved(p))
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
> + Â Â Â Â Â Â Â Â Â Â Â if (!page_count(p)) {
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â if (!PageBuddy(p))
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pos++;
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â else if (PageBuddy(p)) {

just else is okay?

> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â int order = page_order(p);
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pos += (1 << order);
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â }
> + Â Â Â Â Â Â Â Â Â Â Â } else if (PageLRU(p)) {
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â pos++;
> + Â Â Â Â Â Â Â Â Â Â Â } else
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â break;
> + Â Â Â Â Â Â Â }
> + Â Â Â Â Â Â Â spin_unlock_irq(&z->lock);
> + Â Â Â Â Â Â Â if ((pos == pfn + pages) &&
> + Â Â Â Â Â Â Â Â Â Â Â !start_isolate_page_range(pfn, pfn + pages))
> + Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â return pfn;
> + Â Â Â Â Â Â Â if (pos & ((1 << align_order) - 1))
> + Â Â Â Â Â Â Â Â Â Â Â pfn = ALIGN(pos, (1 << align_order));
> + Â Â Â Â Â Â Â else
> + Â Â Â Â Â Â Â Â Â Â Â pfn = pos + (1 << align_order);

pfn has changed here, then why the for loop still need pfn += (1 <<
align_order))?
or maybe I missed something.

> + Â Â Â Â Â Â Â cond_resched();
> + Â Â Â }
> +
> + Â Â Â /* failed */
> + Â Â Â if (blockinfo.end + pages <= end) {
> + Â Â Â Â Â Â Â /* Move base address and find the next block of RAM. */
> + Â Â Â Â Â Â Â base = blockinfo.end;
> + Â Â Â Â Â Â Â goto retry;
> + Â Â Â }
> + Â Â Â return 0;
> +}
>

-- 
Thanks,
--Bob
N‹§²æìr¸›yúèšØb²X¬¶ÇvØ^–)Þ{.nÇ+‰·¥Š{±‘êçzX§¶›¡Ü}©ž²ÆzÚ&j:+v‰¨¾«‘êçzZ+€Ê+zf£¢·hšˆ§~†­†Ûiÿûàz¹®w¥¢¸?™¨è­Ú&¢)ßf”ù^jÇy§m…á@A«a¶Úÿ0¶ìh®å’i