summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorVlastimil Babka <vbabka@suse.cz>2013-09-11 14:22:35 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2013-09-11 15:58:01 -0700
commit7a8010cd36273ff5f6fea5201ef9232f30cebbd9 (patch)
tree3805f3d9a8a1f1c1c555ef31bc1bdb51fb51e33e
parent5b40998ae35cf64561868370e6c9f3d3e94b6bf7 (diff)
downloadop-kernel-dev-7a8010cd36273ff5f6fea5201ef9232f30cebbd9.zip
op-kernel-dev-7a8010cd36273ff5f6fea5201ef9232f30cebbd9.tar.gz
mm: munlock: manual pte walk in fast path instead of follow_page_mask()
Currently munlock_vma_pages_range() calls follow_page_mask() to obtain each individual struct page. This entails repeated full page table translations and page table lock taken for each page separately. This patch avoids the costly follow_page_mask() where possible, by iterating over ptes within single pmd under single page table lock. The first pte is obtained by get_locked_pte() for non-THP page acquired by the initial follow_page_mask(). The rest of the on-stack pagevec for munlock is filled up using pte_walk as long as pte_present() and vm_normal_page() are sufficient to obtain the struct page. After this patch, a 14% speedup was measured for munlocking a 56GB large memory area with THP disabled. Signed-off-by: Vlastimil Babka <vbabka@suse.cz> Cc: Jörn Engel <joern@logfs.org> Cc: Mel Gorman <mgorman@suse.de> Cc: Michel Lespinasse <walken@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Rik van Riel <riel@redhat.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Cc: Vlastimil Babka <vbabka@suse.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
-rw-r--r--include/linux/mm.h12
-rw-r--r--mm/mlock.c110
2 files changed, 85 insertions, 37 deletions
diff --git a/include/linux/mm.h b/include/linux/mm.h
index dce2456..03f84b8 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -643,12 +643,12 @@ static inline enum zone_type page_zonenum(const struct page *page)
#endif
/*
- * The identification function is only used by the buddy allocator for
- * determining if two pages could be buddies. We are not really
- * identifying a zone since we could be using a the section number
- * id if we have not node id available in page flags.
- * We guarantee only that it will return the same value for two
- * combinable pages in a zone.
+ * The identification function is mainly used by the buddy allocator for
+ * determining if two pages could be buddies. We are not really identifying
+ * the zone since we could be using the section number id if we do not have
+ * node id available in page flags.
+ * We only guarantee that it will return the same value for two combinable
+ * pages in a zone.
*/
static inline int page_zone_id(struct page *page)
{
diff --git a/mm/mlock.c b/mm/mlock.c
index 19a934d..d638026 100644
--- a/mm/mlock.c
+++ b/mm/mlock.c
@@ -280,8 +280,7 @@ static void __putback_lru_fast(struct pagevec *pvec, int pgrescued)
* The second phase finishes the munlock only for pages where isolation
* succeeded.
*
- * Note that pvec is modified during the process. Before returning
- * pagevec_reinit() is called on it.
+ * Note that the pagevec may be modified during the process.
*/
static void __munlock_pagevec(struct pagevec *pvec, struct zone *zone)
{
@@ -356,8 +355,60 @@ skip_munlock:
*/
if (pagevec_count(&pvec_putback))
__putback_lru_fast(&pvec_putback, pgrescued);
+}
+
+/*
+ * Fill up pagevec for __munlock_pagevec using pte walk
+ *
+ * The function expects that the struct page corresponding to @start address is
+ * a non-TPH page already pinned and in the @pvec, and that it belongs to @zone.
+ *
+ * The rest of @pvec is filled by subsequent pages within the same pmd and same
+ * zone, as long as the pte's are present and vm_normal_page() succeeds. These
+ * pages also get pinned.
+ *
+ * Returns the address of the next page that should be scanned. This equals
+ * @start + PAGE_SIZE when no page could be added by the pte walk.
+ */
+static unsigned long __munlock_pagevec_fill(struct pagevec *pvec,
+ struct vm_area_struct *vma, int zoneid, unsigned long start,
+ unsigned long end)
+{
+ pte_t *pte;
+ spinlock_t *ptl;
+
+ /*
+ * Initialize pte walk starting at the already pinned page where we
+ * are sure that there is a pte.
+ */
+ pte = get_locked_pte(vma->vm_mm, start, &ptl);
+ end = min(end, pmd_addr_end(start, end));
+
+ /* The page next to the pinned page is the first we will try to get */
+ start += PAGE_SIZE;
+ while (start < end) {
+ struct page *page = NULL;
+ pte++;
+ if (pte_present(*pte))
+ page = vm_normal_page(vma, start, *pte);
+ /*
+ * Break if page could not be obtained or the page's node+zone does not
+ * match
+ */
+ if (!page || page_zone_id(page) != zoneid)
+ break;
- pagevec_reinit(pvec);
+ get_page(page);
+ /*
+ * Increase the address that will be returned *before* the
+ * eventual break due to pvec becoming full by adding the page
+ */
+ start += PAGE_SIZE;
+ if (pagevec_add(pvec, page) == 0)
+ break;
+ }
+ pte_unmap_unlock(pte, ptl);
+ return start;
}
/*
@@ -381,17 +432,16 @@ skip_munlock:
void munlock_vma_pages_range(struct vm_area_struct *vma,
unsigned long start, unsigned long end)
{
- struct pagevec pvec;
- struct zone *zone = NULL;
-
- pagevec_init(&pvec, 0);
vma->vm_flags &= ~VM_LOCKED;
while (start < end) {
- struct page *page;
+ struct page *page = NULL;
unsigned int page_mask, page_increm;
- struct zone *pagezone;
+ struct pagevec pvec;
+ struct zone *zone;
+ int zoneid;
+ pagevec_init(&pvec, 0);
/*
* Although FOLL_DUMP is intended for get_dump_page(),
* it just so happens that its special treatment of the
@@ -400,22 +450,10 @@ void munlock_vma_pages_range(struct vm_area_struct *vma,
* has sneaked into the range, we won't oops here: great).
*/
page = follow_page_mask(vma, start, FOLL_GET | FOLL_DUMP,
- &page_mask);
+ &page_mask);
+
if (page && !IS_ERR(page)) {
- pagezone = page_zone(page);
- /* The whole pagevec must be in the same zone */
- if (pagezone != zone) {
- if (pagevec_count(&pvec))
- __munlock_pagevec(&pvec, zone);
- zone = pagezone;
- }
if (PageTransHuge(page)) {
- /*
- * THP pages are not handled by pagevec due
- * to their possible split (see below).
- */
- if (pagevec_count(&pvec))
- __munlock_pagevec(&pvec, zone);
lock_page(page);
/*
* Any THP page found by follow_page_mask() may
@@ -428,21 +466,31 @@ void munlock_vma_pages_range(struct vm_area_struct *vma,
put_page(page); /* follow_page_mask() */
} else {
/*
- * Non-huge pages are handled in batches
- * via pagevec. The pin from
- * follow_page_mask() prevents them from
- * collapsing by THP.
+ * Non-huge pages are handled in batches via
+ * pagevec. The pin from follow_page_mask()
+ * prevents them from collapsing by THP.
+ */
+ pagevec_add(&pvec, page);
+ zone = page_zone(page);
+ zoneid = page_zone_id(page);
+
+ /*
+ * Try to fill the rest of pagevec using fast
+ * pte walk. This will also update start to
+ * the next page to process. Then munlock the
+ * pagevec.
*/
- if (pagevec_add(&pvec, page) == 0)
- __munlock_pagevec(&pvec, zone);
+ start = __munlock_pagevec_fill(&pvec, vma,
+ zoneid, start, end);
+ __munlock_pagevec(&pvec, zone);
+ goto next;
}
}
page_increm = 1 + (~(start >> PAGE_SHIFT) & page_mask);
start += page_increm * PAGE_SIZE;
+next:
cond_resched();
}
- if (pagevec_count(&pvec))
- __munlock_pagevec(&pvec, zone);
}
/*
OpenPOWER on IntegriCloud