mm, swap: VMA based swap readahead

The swap readahead is an important mechanism to reduce the swap in
latency.  Although pure sequential memory access pattern isn't very
popular for anonymous memory, the space locality is still considered
valid.

In the original swap readahead implementation, the consecutive blocks in
swap device are readahead based on the global space locality estimation.
But the consecutive blocks in swap device just reflect the order of page
reclaiming, don't necessarily reflect the access pattern in virtual
memory.  And the different tasks in the system may have different access
patterns, which makes the global space locality estimation incorrect.

In this patch, when page fault occurs, the virtual pages near the fault
address will be readahead instead of the swap slots near the fault swap
slot in swap device.  This avoid to readahead the unrelated swap slots.
At the same time, the swap readahead is changed to work on per-VMA from
globally.  So that the different access patterns of the different VMAs
could be distinguished, and the different readahead policy could be
applied accordingly.  The original core readahead detection and scaling
algorithm is reused, because it is an effect algorithm to detect the
space locality.

The test and result is as follow,

Common test condition
=====================

Test Machine: Xeon E5 v3 (2 sockets, 72 threads, 32G RAM) Swap device:
NVMe disk

Micro-benchmark with combined access pattern
============================================

vm-scalability, sequential swap test case, 4 processes to eat 50G
virtual memory space, repeat the sequential memory writing until 300
seconds.  The first round writing will trigger swap out, the following
rounds will trigger sequential swap in and out.

At the same time, run vm-scalability random swap test case in
background, 8 processes to eat 30G virtual memory space, repeat the
random memory write until 300 seconds.  This will trigger random swap-in
in the background.

This is a combined workload with sequential and random memory accessing
at the same time.  The result (for sequential workload) is as follow,

			Base		Optimized
			----		---------
throughput		345413 KB/s	414029 KB/s (+19.9%)
latency.average		97.14 us	61.06 us (-37.1%)
latency.50th		2 us		1 us
latency.60th		2 us		1 us
latency.70th		98 us		2 us
latency.80th		160 us		2 us
latency.90th		260 us		217 us
latency.95th		346 us		369 us
latency.99th		1.34 ms		1.09 ms
ra_hit%			52.69%		99.98%

The original swap readahead algorithm is confused by the background
random access workload, so readahead hit rate is lower.  The VMA-base
readahead algorithm works much better.

Linpack
=======

The test memory size is bigger than RAM to trigger swapping.

			Base		Optimized
			----		---------
elapsed_time		393.49 s	329.88 s (-16.2%)
ra_hit%			86.21%		98.82%

The score of base and optimized kernel hasn't visible changes.  But the
elapsed time reduced and readahead hit rate improved, so the optimized
kernel runs better for startup and tear down stages.  And the absolute
value of readahead hit rate is high, shows that the space locality is
still valid in some practical workloads.

Link: http://lkml.kernel.org/r/20170807054038.1843-4-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Shaohua Li <shli@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Tim Chen <tim.c.chen@intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

1 files changed, 198 insertions, 17 deletions

diff --git a/mm/swap_state.c b/mm/swap_state.c
index a901afe9da61..3885fef7bdf5 100644
--- a/mm/swap_state.c
+++ b/mm/swap_state.c
@@ -37,6 +37,29 @@ static const struct address_space_operations swap_aops = {
 
 struct address_space *swapper_spaces[MAX_SWAPFILES];
 static unsigned int nr_swapper_spaces[MAX_SWAPFILES];
+bool swap_vma_readahead = true;
+
+#define SWAP_RA_MAX_ORDER_DEFAULT	3
+
+static int swap_ra_max_order = SWAP_RA_MAX_ORDER_DEFAULT;
+
+#define SWAP_RA_WIN_SHIFT	(PAGE_SHIFT / 2)
+#define SWAP_RA_HITS_MASK	((1UL << SWAP_RA_WIN_SHIFT) - 1)
+#define SWAP_RA_HITS_MAX	SWAP_RA_HITS_MASK
+#define SWAP_RA_WIN_MASK	(~PAGE_MASK & ~SWAP_RA_HITS_MASK)
+
+#define SWAP_RA_HITS(v)		((v) & SWAP_RA_HITS_MASK)
+#define SWAP_RA_WIN(v)		(((v) & SWAP_RA_WIN_MASK) >> SWAP_RA_WIN_SHIFT)
+#define SWAP_RA_ADDR(v)		((v) & PAGE_MASK)
+
+#define SWAP_RA_VAL(addr, win, hits)				\
+	(((addr) & PAGE_MASK) |					\
+	 (((win) << SWAP_RA_WIN_SHIFT) & SWAP_RA_WIN_MASK) |	\
+	 ((hits) & SWAP_RA_HITS_MASK))
+
+/* Initial readahead hits is 4 to start up with a small window */
+#define GET_SWAP_RA_VAL(vma)					\
+	(atomic_long_read(&(vma)->swap_readahead_info) ? : 4)
 
 #define INC_CACHE_INFO(x)	do { swap_cache_info.x++; } while (0)
 #define ADD_CACHE_INFO(x, nr)	do { swap_cache_info.x += (nr); } while (0)
@@ -297,21 +320,36 @@ void free_pages_and_swap_cache(struct page **pages, int nr)
  * lock getting page table operations atomic even if we drop the page
  * lock before returning.
  */
-struct page * lookup_swap_cache(swp_entry_t entry)
+struct page *lookup_swap_cache(swp_entry_t entry, struct vm_area_struct *vma,
+			       unsigned long addr)
 {
 	struct page *page;
+	unsigned long ra_info;
+	int win, hits, readahead;
 
 	page = find_get_page(swap_address_space(entry), swp_offset(entry));
 
-	if (page && likely(!PageTransCompound(page))) {
+	INC_CACHE_INFO(find_total);
+	if (page) {
 		INC_CACHE_INFO(find_success);
-		if (TestClearPageReadahead(page)) {
-			atomic_inc(&swapin_readahead_hits);
+		if (unlikely(PageTransCompound(page)))
+			return page;
+		readahead = TestClearPageReadahead(page);
+		if (vma) {
+			ra_info = GET_SWAP_RA_VAL(vma);
+			win = SWAP_RA_WIN(ra_info);
+			hits = SWAP_RA_HITS(ra_info);
+			if (readahead)
+				hits = min_t(int, hits + 1, SWAP_RA_HITS_MAX);
+			atomic_long_set(&vma->swap_readahead_info,
+					SWAP_RA_VAL(addr, win, hits));
+		}
+		if (readahead) {
 			count_vm_event(SWAP_RA_HIT);
+			if (!vma)
+				atomic_inc(&swapin_readahead_hits);
 		}
 	}
-
-	INC_CACHE_INFO(find_total);
 	return page;
 }
 
@@ -426,22 +464,20 @@ struct page *read_swap_cache_async(swp_entry_t entry, gfp_t gfp_mask,
 	return retpage;
 }
 
-static unsigned long swapin_nr_pages(unsigned long offset)
+static unsigned int __swapin_nr_pages(unsigned long prev_offset,
+				      unsigned long offset,
+				      int hits,
+				      int max_pages,
+				      int prev_win)
 {
-	static unsigned long prev_offset;
-	unsigned int pages, max_pages, last_ra;
-	static atomic_t last_readahead_pages;
-
-	max_pages = 1 << READ_ONCE(page_cluster);
-	if (max_pages <= 1)
-		return 1;
+	unsigned int pages, last_ra;
 
 	/*
 	 * This heuristic has been found to work well on both sequential and
 	 * random loads, swapping to hard disk or to SSD: please don't ask
 	 * what the "+ 2" means, it just happens to work well, that's all.
 	 */
-	pages = atomic_xchg(&swapin_readahead_hits, 0) + 2;
+	pages = hits + 2;
 	if (pages == 2) {
 		/*
 		 * We can have no readahead hits to judge by: but must not get
@@ -450,7 +486,6 @@ static unsigned long swapin_nr_pages(unsigned long offset)
 		 */
 		if (offset != prev_offset + 1 && offset != prev_offset - 1)
 			pages = 1;
-		prev_offset = offset;
 	} else {
 		unsigned int roundup = 4;
 		while (roundup < pages)
@@ -462,9 +497,28 @@ static unsigned long swapin_nr_pages(unsigned long offset)
 		pages = max_pages;
 
 	/* Don't shrink readahead too fast */
-	last_ra = atomic_read(&last_readahead_pages) / 2;
+	last_ra = prev_win / 2;
 	if (pages < last_ra)
 		pages = last_ra;
+
+	return pages;
+}
+
+static unsigned long swapin_nr_pages(unsigned long offset)
+{
+	static unsigned long prev_offset;
+	unsigned int hits, pages, max_pages;
+	static atomic_t last_readahead_pages;
+
+	max_pages = 1 << READ_ONCE(page_cluster);
+	if (max_pages <= 1)
+		return 1;
+
+	hits = atomic_xchg(&swapin_readahead_hits, 0);
+	pages = __swapin_nr_pages(prev_offset, offset, hits, max_pages,
+				  atomic_read(&last_readahead_pages));
+	if (!hits)
+		prev_offset = offset;
 	atomic_set(&last_readahead_pages, pages);
 
 	return pages;
@@ -570,3 +624,130 @@ void exit_swap_address_space(unsigned int type)
 	synchronize_rcu();
 	kvfree(spaces);
 }
+
+static inline void swap_ra_clamp_pfn(struct vm_area_struct *vma,
+				     unsigned long faddr,
+				     unsigned long lpfn,
+				     unsigned long rpfn,
+				     unsigned long *start,
+				     unsigned long *end)
+{
+	*start = max3(lpfn, PFN_DOWN(vma->vm_start),
+		      PFN_DOWN(faddr & PMD_MASK));
+	*end = min3(rpfn, PFN_DOWN(vma->vm_end),
+		    PFN_DOWN((faddr & PMD_MASK) + PMD_SIZE));
+}
+
+struct page *swap_readahead_detect(struct vm_fault *vmf,
+				   struct vma_swap_readahead *swap_ra)
+{
+	struct vm_area_struct *vma = vmf->vma;
+	unsigned long swap_ra_info;
+	struct page *page;
+	swp_entry_t entry;
+	unsigned long faddr, pfn, fpfn;
+	unsigned long start, end;
+	pte_t *pte;
+	unsigned int max_win, hits, prev_win, win, left;
+#ifndef CONFIG_64BIT
+	pte_t *tpte;
+#endif
+
+	faddr = vmf->address;
+	entry = pte_to_swp_entry(vmf->orig_pte);
+	if ((unlikely(non_swap_entry(entry))))
+		return NULL;
+	page = lookup_swap_cache(entry, vma, faddr);
+	if (page)
+		return page;
+
+	max_win = 1 << READ_ONCE(swap_ra_max_order);
+	if (max_win == 1) {
+		swap_ra->win = 1;
+		return NULL;
+	}
+
+	fpfn = PFN_DOWN(faddr);
+	swap_ra_info = GET_SWAP_RA_VAL(vma);
+	pfn = PFN_DOWN(SWAP_RA_ADDR(swap_ra_info));
+	prev_win = SWAP_RA_WIN(swap_ra_info);
+	hits = SWAP_RA_HITS(swap_ra_info);
+	swap_ra->win = win = __swapin_nr_pages(pfn, fpfn, hits,
+					       max_win, prev_win);
+	atomic_long_set(&vma->swap_readahead_info,
+			SWAP_RA_VAL(faddr, win, 0));
+
+	if (win == 1)
+		return NULL;
+
+	/* Copy the PTEs because the page table may be unmapped */
+	if (fpfn == pfn + 1)
+		swap_ra_clamp_pfn(vma, faddr, fpfn, fpfn + win, &start, &end);
+	else if (pfn == fpfn + 1)
+		swap_ra_clamp_pfn(vma, faddr, fpfn - win + 1, fpfn + 1,
+				  &start, &end);
+	else {
+		left = (win - 1) / 2;
+		swap_ra_clamp_pfn(vma, faddr, fpfn - left, fpfn + win - left,
+				  &start, &end);
+	}
+	swap_ra->nr_pte = end - start;
+	swap_ra->offset = fpfn - start;
+	pte = vmf->pte - swap_ra->offset;
+#ifdef CONFIG_64BIT
+	swap_ra->ptes = pte;
+#else
+	tpte = swap_ra->ptes;
+	for (pfn = start; pfn != end; pfn++)
+		*tpte++ = *pte++;
+#endif
+
+	return NULL;
+}
+
+struct page *do_swap_page_readahead(swp_entry_t fentry, gfp_t gfp_mask,
+				    struct vm_fault *vmf,
+				    struct vma_swap_readahead *swap_ra)
+{
+	struct blk_plug plug;
+	struct vm_area_struct *vma = vmf->vma;
+	struct page *page;
+	pte_t *pte, pentry;
+	swp_entry_t entry;
+	unsigned int i;
+	bool page_allocated;
+
+	if (swap_ra->win == 1)
+		goto skip;
+
+	blk_start_plug(&plug);
+	for (i = 0, pte = swap_ra->ptes; i < swap_ra->nr_pte;
+	     i++, pte++) {
+		pentry = *pte;
+		if (pte_none(pentry))
+			continue;
+		if (pte_present(pentry))
+			continue;
+		entry = pte_to_swp_entry(pentry);
+		if (unlikely(non_swap_entry(entry)))
+			continue;
+		page = __read_swap_cache_async(entry, gfp_mask, vma,
+					       vmf->address, &page_allocated);
+		if (!page)
+			continue;
+		if (page_allocated) {
+			swap_readpage(page, false);
+			if (i != swap_ra->offset &&
+			    likely(!PageTransCompound(page))) {
+				SetPageReadahead(page);
+				count_vm_event(SWAP_RA);
+			}
+		}
+		put_page(page);
+	}
+	blk_finish_plug(&plug);
+	lru_add_drain();
+skip:
+	return read_swap_cache_async(fentry, gfp_mask, vma, vmf->address,
+				     swap_ra->win == 1);
+}