[mm,v5,4/7] mm: Initialize MAX_ORDER_NR_PAGES at a time instead of doing larger sections
diff mbox series

Message ID 154145278583.30046.4918131143612801028.stgit@ahduyck-desk1.jf.intel.com
State Superseded
Headers show
Series
  • Deferred page init improvements
Related show

Commit Message

Alexander Duyck Nov. 5, 2018, 9:19 p.m. UTC
This patch adds yet another iterator, for_each_free_mem_range_in_zone_from.
It and then uses it to support initializing and freeing pages in groups no
larger than MAX_ORDER_NR_PAGES. By doing this we can greatly improve the
cache locality of the pages while we do several loops over them in the init
and freeing process.

We are able to tighten the loops as a result since we only really need the
checks for first_init_pfn in our first iteration and after that we can
assume that all future values will be greater than this. So I have added a
function called deferred_init_mem_pfn_range_in_zone that primes the
iterators and if it fails we can just exit.

On my x86_64 test system with 384GB of memory per node I saw a reduction in
initialization time from 1.85s to 1.38s as a result of this patch.

Signed-off-by: Alexander Duyck <alexander.h.duyck@linux.intel.com>
---
 include/linux/memblock.h |   16 +++++
 mm/page_alloc.c          |  162 ++++++++++++++++++++++++++++++++++------------
 2 files changed, 134 insertions(+), 44 deletions(-)

Comments

Pavel Tatashin Nov. 10, 2018, 1:02 a.m. UTC | #1
On 18-11-05 13:19:45, Alexander Duyck wrote:
>  	}
> -	first_init_pfn = max(zone->zone_start_pfn, first_init_pfn);
> +
> +	/* If the zone is empty somebody else may have cleared out the zone */
> +	if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
> +						 first_init_pfn)) {
> +		pgdat_resize_unlock(pgdat, &flags);
> +		pgdat_init_report_one_done();
> +		return 0;

It would make more sense to add goto to the end of this function and report
that in Node X had 0 pages initialized. Otherwise, it will be confusing
why some nodes are not initialized during boot. It is simply because
they do not have deferred pages left to be initialized.


> +	}
>  
>  	/*
> -	 * Initialize and free pages. We do it in two loops: first we initialize
> -	 * struct page, than free to buddy allocator, because while we are
> -	 * freeing pages we can access pages that are ahead (computing buddy
> -	 * page in __free_one_page()).
> +	 * Initialize and free pages in MAX_ORDER sized increments so
> +	 * that we can avoid introducing any issues with the buddy
> +	 * allocator.
>  	 */
> -	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
> -		spfn = max_t(unsigned long, first_init_pfn, spfn);
> -		nr_pages += deferred_init_pages(zone, spfn, epfn);
> -	}
> -	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
> -		spfn = max_t(unsigned long, first_init_pfn, spfn);
> -		deferred_free_pages(spfn, epfn);
> -	}
> +	while (spfn < epfn)
> +		nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
> +
>  	pgdat_resize_unlock(pgdat, &flags);
>  
>  	/* Sanity check that the next zone really is unpopulated */
> @@ -1602,9 +1689,9 @@ deferred_grow_zone(struct zone *zone, unsigned int order)
>  {

How about order = max(order, max_order)?

>  	unsigned long nr_pages_needed = ALIGN(1 << order, PAGES_PER_SECTION);


> -	first_init_pfn = max(zone->zone_start_pfn, first_deferred_pfn);
> -
> -	if (first_init_pfn >= pgdat_end_pfn(pgdat)) {
> +	/* If the zone is empty somebody else may have cleared out the zone */
> +	if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
> +						 first_deferred_pfn)) {
>  		pgdat_resize_unlock(pgdat, &flags);
> -		return false;
> +		return true;

I am OK to change to true here, please also set
pgdat->first_deferred_pfn to ULONG_MAX to indicate that there are no
more deferred pages in this node.


Overall, I like this patch, makes things a lot easier, assuming the
above is addressed:

Reviewed-by: Pavel Tatashin <pasha.tatashin@soleen.com>

Thank you,
Pasha
Alexander Duyck Nov. 19, 2018, 6:53 p.m. UTC | #2
On Fri, 2018-11-09 at 20:02 -0500, Pavel Tatashin wrote:
> On 18-11-05 13:19:45, Alexander Duyck wrote:
> >  	}
> > -	first_init_pfn = max(zone->zone_start_pfn, first_init_pfn);
> > +
> > +	/* If the zone is empty somebody else may have cleared out the zone */
> > +	if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
> > +						 first_init_pfn)) {
> > +		pgdat_resize_unlock(pgdat, &flags);
> > +		pgdat_init_report_one_done();
> > +		return 0;
> 
> It would make more sense to add goto to the end of this function and report
> that in Node X had 0 pages initialized. Otherwise, it will be confusing
> why some nodes are not initialized during boot. It is simply because
> they do not have deferred pages left to be initialized.

This part makes sense and will be implemented for v6.

> 
> > +	}
> >  
> >  	/*
> > -	 * Initialize and free pages. We do it in two loops: first we initialize
> > -	 * struct page, than free to buddy allocator, because while we are
> > -	 * freeing pages we can access pages that are ahead (computing buddy
> > -	 * page in __free_one_page()).
> > +	 * Initialize and free pages in MAX_ORDER sized increments so
> > +	 * that we can avoid introducing any issues with the buddy
> > +	 * allocator.
> >  	 */
> > -	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
> > -		spfn = max_t(unsigned long, first_init_pfn, spfn);
> > -		nr_pages += deferred_init_pages(zone, spfn, epfn);
> > -	}
> > -	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
> > -		spfn = max_t(unsigned long, first_init_pfn, spfn);
> > -		deferred_free_pages(spfn, epfn);
> > -	}
> > +	while (spfn < epfn)
> > +		nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
> > +
> >  	pgdat_resize_unlock(pgdat, &flags);
> >  
> >  	/* Sanity check that the next zone really is unpopulated */
> > @@ -1602,9 +1689,9 @@ deferred_grow_zone(struct zone *zone, unsigned int order)
> >  {
> 
> How about order = max(order, max_order)?

I'm not sure what you are getting at here? The "order" variable passed
is only really used to get the nr_pages_needed, and that value is
rounded up to PAGES_PER_SECTION which should always be equal to or
greater than MAX_ORDER pages.

If order is greater than MAX_ORDER we would want to process some number
of MAX_ORDER sized blocks until we get to the size specified by order.
If we process it all as one large block it will perform worse as we
risk pushing page structs out of the cache.

> >  	unsigned long nr_pages_needed = ALIGN(1 << order, PAGES_PER_SECTION);
> 
> 
> > -	first_init_pfn = max(zone->zone_start_pfn, first_deferred_pfn);
> > -
> > -	if (first_init_pfn >= pgdat_end_pfn(pgdat)) {
> > +	/* If the zone is empty somebody else may have cleared out the zone */
> > +	if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
> > +						 first_deferred_pfn)) {
> >  		pgdat_resize_unlock(pgdat, &flags);
> > -		return false;
> > +		return true;
> 
> I am OK to change to true here, please also set
> pgdat->first_deferred_pfn to ULONG_MAX to indicate that there are no
> more deferred pages in this node.

Done for v6 of the patch set.

> 
> Overall, I like this patch, makes things a lot easier, assuming the
> above is addressed:
> 
> Reviewed-by: Pavel Tatashin <pasha.tatashin@soleen.com>
> 
> Thank you,
> Pasha

Patch
diff mbox series

diff --git a/include/linux/memblock.h b/include/linux/memblock.h
index 413623dc96a3..5ba52a7878a0 100644
--- a/include/linux/memblock.h
+++ b/include/linux/memblock.h
@@ -268,6 +268,22 @@  void __next_mem_pfn_range_in_zone(u64 *idx, struct zone *zone,
 	     __next_mem_pfn_range_in_zone(&i, zone, p_start, p_end);	\
 	     i != (u64)ULLONG_MAX;					\
 	     __next_mem_pfn_range_in_zone(&i, zone, p_start, p_end))
+
+/**
+ * for_each_free_mem_range_in_zone_from - iterate through zone specific
+ * free memblock areas from a given point
+ * @i: u64 used as loop variable
+ * @zone: zone in which all of the memory blocks reside
+ * @p_start: ptr to phys_addr_t for start address of the range, can be %NULL
+ * @p_end: ptr to phys_addr_t for end address of the range, can be %NULL
+ *
+ * Walks over free (memory && !reserved) areas of memblock in a specific
+ * zone, continuing from current position. Available as soon as memblock is
+ * initialized.
+ */
+#define for_each_free_mem_pfn_range_in_zone_from(i, zone, p_start, p_end) \
+	for (; i != (u64)ULLONG_MAX;					  \
+	     __next_mem_pfn_range_in_zone(&i, zone, p_start, p_end))
 #endif /* CONFIG_DEFERRED_STRUCT_PAGE_INIT */
 
 /**
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 5cfd3ebe10d1..3466a01ed90a 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1512,16 +1512,102 @@  static unsigned long  __init deferred_init_pages(struct zone *zone,
 	return (nr_pages);
 }
 
+/*
+ * This function is meant to pre-load the iterator for the zone init.
+ * Specifically it walks through the ranges until we are caught up to the
+ * first_init_pfn value and exits there. If we never encounter the value we
+ * return false indicating there are no valid ranges left.
+ */
+static bool __init
+deferred_init_mem_pfn_range_in_zone(u64 *i, struct zone *zone,
+				    unsigned long *spfn, unsigned long *epfn,
+				    unsigned long first_init_pfn)
+{
+	u64 j;
+
+	/*
+	 * Start out by walking through the ranges in this zone that have
+	 * already been initialized. We don't need to do anything with them
+	 * so we just need to flush them out of the system.
+	 */
+	for_each_free_mem_pfn_range_in_zone(j, zone, spfn, epfn) {
+		if (*epfn <= first_init_pfn)
+			continue;
+		if (*spfn < first_init_pfn)
+			*spfn = first_init_pfn;
+		*i = j;
+		return true;
+	}
+
+	return false;
+}
+
+/*
+ * Initialize and free pages. We do it in two loops: first we initialize
+ * struct page, than free to buddy allocator, because while we are
+ * freeing pages we can access pages that are ahead (computing buddy
+ * page in __free_one_page()).
+ *
+ * In order to try and keep some memory in the cache we have the loop
+ * broken along max page order boundaries. This way we will not cause
+ * any issues with the buddy page computation.
+ */
+static unsigned long __init
+deferred_init_maxorder(u64 *i, struct zone *zone, unsigned long *start_pfn,
+		       unsigned long *end_pfn)
+{
+	unsigned long mo_pfn = ALIGN(*start_pfn + 1, MAX_ORDER_NR_PAGES);
+	unsigned long spfn = *start_pfn, epfn = *end_pfn;
+	unsigned long nr_pages = 0;
+	u64 j = *i;
+
+	/* First we loop through and initialize the page values */
+	for_each_free_mem_pfn_range_in_zone_from(j, zone, &spfn, &epfn) {
+		unsigned long t;
+
+		if (mo_pfn <= spfn)
+			break;
+
+		t = min(mo_pfn, epfn);
+		nr_pages += deferred_init_pages(zone, spfn, t);
+
+		if (mo_pfn <= epfn)
+			break;
+	}
+
+	/* Reset values and now loop through freeing pages as needed */
+	j = *i;
+
+	for_each_free_mem_pfn_range_in_zone_from(j, zone, start_pfn, end_pfn) {
+		unsigned long t;
+
+		if (mo_pfn <= *start_pfn)
+			break;
+
+		t = min(mo_pfn, *end_pfn);
+		deferred_free_pages(*start_pfn, t);
+		*start_pfn = t;
+
+		if (mo_pfn < *end_pfn)
+			break;
+	}
+
+	/* Store our current values to be reused on the next iteration */
+	*i = j;
+
+	return nr_pages;
+}
+
 /* Initialise remaining memory on a node */
 static int __init deferred_init_memmap(void *data)
 {
 	pg_data_t *pgdat = data;
+	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
+	unsigned long spfn = 0, epfn = 0, nr_pages = 0;
+	unsigned long first_init_pfn, flags;
 	unsigned long start = jiffies;
-	unsigned long nr_pages = 0;
-	unsigned long spfn, epfn, first_init_pfn, flags;
-	int zid;
 	struct zone *zone;
-	const struct cpumask *cpumask = cpumask_of_node(pgdat->node_id);
+	int zid;
 	u64 i;
 
 	/* Bind memory initialisation thread to a local node if possible */
@@ -1547,22 +1633,23 @@  static int __init deferred_init_memmap(void *data)
 		if (first_init_pfn < zone_end_pfn(zone))
 			break;
 	}
-	first_init_pfn = max(zone->zone_start_pfn, first_init_pfn);
+
+	/* If the zone is empty somebody else may have cleared out the zone */
+	if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
+						 first_init_pfn)) {
+		pgdat_resize_unlock(pgdat, &flags);
+		pgdat_init_report_one_done();
+		return 0;
+	}
 
 	/*
-	 * Initialize and free pages. We do it in two loops: first we initialize
-	 * struct page, than free to buddy allocator, because while we are
-	 * freeing pages we can access pages that are ahead (computing buddy
-	 * page in __free_one_page()).
+	 * Initialize and free pages in MAX_ORDER sized increments so
+	 * that we can avoid introducing any issues with the buddy
+	 * allocator.
 	 */
-	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
-		spfn = max_t(unsigned long, first_init_pfn, spfn);
-		nr_pages += deferred_init_pages(zone, spfn, epfn);
-	}
-	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
-		spfn = max_t(unsigned long, first_init_pfn, spfn);
-		deferred_free_pages(spfn, epfn);
-	}
+	while (spfn < epfn)
+		nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
+
 	pgdat_resize_unlock(pgdat, &flags);
 
 	/* Sanity check that the next zone really is unpopulated */
@@ -1602,9 +1689,9 @@  deferred_grow_zone(struct zone *zone, unsigned int order)
 {
 	unsigned long nr_pages_needed = ALIGN(1 << order, PAGES_PER_SECTION);
 	pg_data_t *pgdat = zone->zone_pgdat;
-	unsigned long nr_pages = 0;
-	unsigned long first_init_pfn, spfn, epfn, t, flags;
 	unsigned long first_deferred_pfn = pgdat->first_deferred_pfn;
+	unsigned long spfn, epfn, flags;
+	unsigned long nr_pages = 0;
 	u64 i;
 
 	/* Only the last zone may have deferred pages */
@@ -1633,36 +1720,23 @@  deferred_grow_zone(struct zone *zone, unsigned int order)
 		return true;
 	}
 
-	first_init_pfn = max(zone->zone_start_pfn, first_deferred_pfn);
-
-	if (first_init_pfn >= pgdat_end_pfn(pgdat)) {
+	/* If the zone is empty somebody else may have cleared out the zone */
+	if (!deferred_init_mem_pfn_range_in_zone(&i, zone, &spfn, &epfn,
+						 first_deferred_pfn)) {
 		pgdat_resize_unlock(pgdat, &flags);
-		return false;
+		return true;
 	}
 
-	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
-		spfn = max_t(unsigned long, first_init_pfn, spfn);
-
-		while (spfn < epfn && nr_pages < nr_pages_needed) {
-			t = ALIGN(spfn + PAGES_PER_SECTION, PAGES_PER_SECTION);
-			first_deferred_pfn = min(t, epfn);
-			nr_pages += deferred_init_pages(zone, spfn,
-							first_deferred_pfn);
-			spfn = first_deferred_pfn;
-		}
-
-		if (nr_pages >= nr_pages_needed)
-			break;
+	/*
+	 * Initialize and free pages in MAX_ORDER sized increments so
+	 * that we can avoid introducing any issues with the buddy
+	 * allocator.
+	 */
+	while (spfn < epfn && nr_pages < nr_pages_needed) {
+		nr_pages += deferred_init_maxorder(&i, zone, &spfn, &epfn);
+		first_deferred_pfn = spfn;
 	}
 
-	for_each_free_mem_pfn_range_in_zone(i, zone, &spfn, &epfn) {
-		spfn = max_t(unsigned long, first_init_pfn, spfn);
-		epfn = min_t(unsigned long, first_deferred_pfn, epfn);
-		deferred_free_pages(spfn, epfn);
-
-		if (first_deferred_pfn == epfn)
-			break;
-	}
 	pgdat->first_deferred_pfn = first_deferred_pfn;
 	pgdat_resize_unlock(pgdat, &flags);