Skip to content

Commit

Permalink
mm, THP, swap: delay splitting THP after swapped out
Browse files Browse the repository at this point in the history 
In this patch, splitting transparent huge page (THP) during swapping out
is delayed from after adding the THP into the swap cache to after
swapping out finishes.  After the patch, more operations for the
anonymous THP reclaiming, such as writing the THP to the swap device,
removing the THP from the swap cache could be batched.  So that the
performance of anonymous THP swapping out could be improved.

This is the second step for the THP swap support.  The plan is to delay
splitting the THP step by step and avoid splitting the THP finally.

With the patchset, the swap out throughput improves 42% (from about
5.81GB/s to about 8.25GB/s) in the vm-scalability swap-w-seq test case
with 16 processes.  At the same time, the IPI (reflect TLB flushing)
reduced about 78.9%.  The test is done on a Xeon E5 v3 system.  The swap
device used is a RAM simulated PMEM (persistent memory) device.  To test
the sequential swapping out, the test case creates 8 processes, which
sequentially allocate and write to the anonymous pages until the RAM and
part of the swap device is used up.

Link: http://lkml.kernel.org/r/[email protected]
Signed-off-by: "Huang, Ying" <[email protected]>
Cc: Johannes Weiner <[email protected]>
Cc: Minchan Kim <[email protected]>
Cc: Hugh Dickins <[email protected]>
Cc: Shaohua Li <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: Andrea Arcangeli <[email protected]>
Cc: "Kirill A . Shutemov" <[email protected]>
Cc: Michal Hocko <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: Jens Axboe <[email protected]>
Cc: Ross Zwisler <[email protected]> [for brd.c, zram_drv.c, pmem.c]
Cc: Vishal L Verma <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
  • Loading branch information
@yhuang-intel @torvalds
yhuang-intel authored and torvalds committed Sep 7, 2017
1 parent d6810d7 commit bd4c82c
Showing 1 changed file with 52 additions and 43 deletions.
95 changes: 52 additions & 43 deletions mm/vmscan.c
View file
Original file line number Diff line number Diff line change
Expand Up @@ -536,7 +536,9 @@ static inline int is_page_cache_freeable(struct page *page)
* that isolated the page, the page cache radix tree and
* optional buffer heads at page->private.
*/
return page_count(page) - page_has_private(page) == 2;
int radix_pins = PageTransHuge(page) && PageSwapCache(page) ?
HPAGE_PMD_NR : 1;
return page_count(page) - page_has_private(page) == 1 + radix_pins;
}

static int may_write_to_inode(struct inode *inode, struct scan_control *sc)
Expand Down Expand Up @@ -666,6 +668,7 @@ static int __remove_mapping(struct address_space *mapping, struct page *page,
bool reclaimed)
{
unsigned long flags;
int refcount;

BUG_ON(!PageLocked(page));
BUG_ON(mapping != page_mapping(page));
Expand Down Expand Up @@ -696,11 +699,15 @@ static int __remove_mapping(struct address_space *mapping, struct page *page,
* Note that if SetPageDirty is always performed via set_page_dirty,
* and thus under tree_lock, then this ordering is not required.
*/
if (!page_ref_freeze(page, 2))
if (unlikely(PageTransHuge(page)) && PageSwapCache(page))
refcount = 1 + HPAGE_PMD_NR;
else
refcount = 2;
if (!page_ref_freeze(page, refcount))
goto cannot_free;
/* note: atomic_cmpxchg in page_freeze_refs provides the smp_rmb */
if (unlikely(PageDirty(page))) {
page_ref_unfreeze(page, 2);
page_ref_unfreeze(page, refcount);
goto cannot_free;
}

Expand Down Expand Up @@ -1122,58 +1129,56 @@ static unsigned long shrink_page_list(struct list_head *page_list,
* Try to allocate it some swap space here.
* Lazyfree page could be freed directly
*/
if (PageAnon(page) && PageSwapBacked(page) &&
!PageSwapCache(page)) {
if (!(sc->gfp_mask & __GFP_IO))
goto keep_locked;
if (PageTransHuge(page)) {
/* cannot split THP, skip it */
if (!can_split_huge_page(page, NULL))
goto activate_locked;
/*
* Split pages without a PMD map right
* away. Chances are some or all of the
* tail pages can be freed without IO.
*/
if (!compound_mapcount(page) &&
split_huge_page_to_list(page, page_list))
goto activate_locked;
}
if (!add_to_swap(page)) {
if (!PageTransHuge(page))
goto activate_locked;
/* Split THP and swap individual base pages */
if (split_huge_page_to_list(page, page_list))
goto activate_locked;
if (!add_to_swap(page))
goto activate_locked;
}

/* XXX: We don't support THP writes */
if (PageTransHuge(page) &&
split_huge_page_to_list(page, page_list)) {
delete_from_swap_cache(page);
goto activate_locked;
}
if (PageAnon(page) && PageSwapBacked(page)) {
if (!PageSwapCache(page)) {
if (!(sc->gfp_mask & __GFP_IO))
goto keep_locked;
if (PageTransHuge(page)) {
/* cannot split THP, skip it */
if (!can_split_huge_page(page, NULL))
goto activate_locked;
/*
* Split pages without a PMD map right
* away. Chances are some or all of the
* tail pages can be freed without IO.
*/
if (!compound_mapcount(page) &&
split_huge_page_to_list(page,
page_list))
goto activate_locked;
}
if (!add_to_swap(page)) {
if (!PageTransHuge(page))
goto activate_locked;
/* Fallback to swap normal pages */
if (split_huge_page_to_list(page,
page_list))
goto activate_locked;
if (!add_to_swap(page))
goto activate_locked;
}

may_enter_fs = 1;
may_enter_fs = 1;

/* Adding to swap updated mapping */
mapping = page_mapping(page);
/* Adding to swap updated mapping */
mapping = page_mapping(page);
}
} else if (unlikely(PageTransHuge(page))) {
/* Split file THP */
if (split_huge_page_to_list(page, page_list))
goto keep_locked;
}

VM_BUG_ON_PAGE(PageTransHuge(page), page);

/*
* The page is mapped into the page tables of one or more
* processes. Try to unmap it here.
*/
if (page_mapped(page)) {
if (!try_to_unmap(page, ttu_flags | TTU_BATCH_FLUSH)) {
enum ttu_flags flags = ttu_flags | TTU_BATCH_FLUSH;

if (unlikely(PageTransHuge(page)))
flags |= TTU_SPLIT_HUGE_PMD;
if (!try_to_unmap(page, flags)) {
nr_unmap_fail++;
goto activate_locked;
}
Expand Down Expand Up @@ -1313,7 +1318,11 @@ static unsigned long shrink_page_list(struct list_head *page_list,
* Is there need to periodically free_page_list? It would
* appear not as the counts should be low
*/
list_add(&page->lru, &free_pages);
if (unlikely(PageTransHuge(page))) {
mem_cgroup_uncharge(page);
(*get_compound_page_dtor(page))(page);
} else
list_add(&page->lru, &free_pages);
continue;

activate_locked:
Expand Down

0 comments on commit bd4c82c

Please sign in to comment.