Fix the incorrect memmep defer init handling and do some cleanup

VMware observed a performance regression during memmap init on their platform, and bisected to commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN") causing it.

Before the commit:

  [0.033176] Normal zone: 1445888 pages used for memmap
  [0.033176] Normal zone: 89391104 pages, LIFO batch:63
  [0.035851] ACPI: PM-Timer IO Port: 0x448

With commit

  [0.026874] Normal zone: 1445888 pages used for memmap
  [0.026875] Normal zone: 89391104 pages, LIFO batch:63
  [2.028450] ACPI: PM-Timer IO Port: 0x448

The root cause is the current memmap defer init doesn't work as expected. Before, memmap_init_zone() was used to do memmap init of one whole zone, to initialize all low zones of one numa node, but defer memmap init of the last zone in that numa node. However, since commit 73a6e474cb376, function memmap_init() is adapted to iterater over memblock regions inside one zone, then call memmap_init_zone() to do memmap init for each region.

E.g, on VMware's system, the memory layout is as below, there are two memory regions in node 2. The current code will mistakenly initialize the whole 1st region [mem 0xab00000000-0xfcffffffff], then do memmap defer to iniatialize only one memmory section on the 2nd region [mem 0x10000000000-0x1033fffffff]. In fact, we only expect to see that there's only one memory section's memmap initialized. That's why more time is costed at the time.

[    0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00000000-0x0009ffff]
[    0.008842] ACPI: SRAT: Node 0 PXM 0 [mem 0x00100000-0xbfffffff]
[    0.008843] ACPI: SRAT: Node 0 PXM 0 [mem 0x100000000-0x55ffffffff]
[    0.008844] ACPI: SRAT: Node 1 PXM 1 [mem 0x5600000000-0xaaffffffff]
[    0.008844] ACPI: SRAT: Node 2 PXM 2 [mem 0xab00000000-0xfcffffffff]
[    0.008845] ACPI: SRAT: Node 2 PXM 2 [mem 0x10000000000-0x1033fffffff]

Now, let's add a parameter 'zone_end_pfn' to memmap_init_zone() to pass down the real zone end pfn so that defer_init() can use it to judge whether defer need be taken in zone wide.

Fixes commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN")

Reported-by Rahul Gopakumar <[email protected]>

Signed-off-by Baoquan He <[email protected]>

Cc [email protected]

---
 arch/ia64/mm/init.c | 4 ++--
 include/linux/mm.h  | 5 +++--
 mm/memory_hotplug.c | 2 +-
 mm/page_alloc.c     | 8 +++++---
 4 files changed, 11 insertions(+), 8 deletions(-)

arch/ia64/mm/init.c +2 -2

index 9b5acf8fb092..e76386a3479e 100644
@@ -536,7 +536,7 @@ virtual_memmap_init(u64 start, u64 end, void *arg)
 
 	if (map_start < map_end)
 		memmap_init_zone((unsigned long)(map_end - map_start),
-				 args->nid, args->zone, page_to_pfn(map_start),
+				 args->nid, args->zone, page_to_pfn(map_start), page_to_pfn(map_end),
 				 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 	return 0;
 }

@@ -546,7 +546,7 @@ memmap_init (unsigned long size, int nid, unsigned long zone,
 	     unsigned long start_pfn)
 {
 	if (!vmem_map) {
-		memmap_init_zone(size, nid, zone, start_pfn,
+		memmap_init_zone(size, nid, zone, start_pfn, start_pfn + size,
 				 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 	} else {
 		struct page *start;

include/linux/mm.h +3 -2

index e4e5be20b0c2..92e06ea053f4 100644
@@ -2432,8 +2432,9 @@ extern int __meminit early_pfn_to_nid(unsigned long pfn);
 #endif
 
 extern void set_dma_reserve(unsigned long new_dma_reserve);
-extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
-		enum meminit_context, struct vmem_altmap *, int migratetype);
+extern void memmap_init_zone(unsigned long, int, unsigned long,
+		unsigned long, unsigned long, enum meminit_context,
+		struct vmem_altmap *, int migratetype);
 extern void setup_per_zone_wmarks(void);
 extern int __meminit init_per_zone_wmark_min(void);
 extern void mem_init(void);

mm/memory_hotplug.c +1 -1

index af41fb990820..f9d57b9be8c7 100644
@@ -713,7 +713,7 @@ void __ref move_pfn_range_to_zone(struct zone *zone, unsigned long start_pfn,
 	 * expects the zone spans the pfn range. All the pages in the range
 	 * are reserved so nobody should be touching them so we should be safe
 	 */
-	memmap_init_zone(nr_pages, nid, zone_idx(zone), start_pfn,
+	memmap_init_zone(nr_pages, nid, zone_idx(zone), start_pfn, 0,
 			 MEMINIT_HOTPLUG, altmap, migratetype);
 
 	set_zone_contiguous(zone);

mm/page_alloc.c +5 -3

index 8cea0823b70e..32645f2e7b96 100644
@@ -423,6 +423,8 @@ defer_init(int nid, unsigned long pfn, unsigned long end_pfn)
 	if (end_pfn < pgdat_end_pfn(NODE_DATA(nid)))
 		return false;
 
+	if (NODE_DATA(nid)->first_deferred_pfn != ULONG_MAX)
+		return true;
 	/*
 	 * We start only with one section of pages, more pages are added as
 	 * needed until the rest of deferred pages are initialized.

@@ -6116,7 +6118,7 @@ overlap_memmap_init(unsigned long zone, unsigned long *pfn)
  * zone stats (e.g., nr_isolate_pageblock) are touched.
  */
 void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
-		unsigned long start_pfn,
+		unsigned long start_pfn, unsigned long zone_end_pfn,
 		enum meminit_context context,
 		struct vmem_altmap *altmap, int migratetype)
 {

@@ -6152,7 +6154,7 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
 		if (context == MEMINIT_EARLY) {
 			if (overlap_memmap_init(zone, &pfn))
 				continue;
-			if (defer_init(nid, pfn, end_pfn))
+			if (defer_init(nid, pfn, zone_end_pfn))
 				break;
 		}
 

@@ -6307,7 +6309,7 @@ void __init __weak memmap_init(unsigned long size, int nid,
 
 		if (end_pfn > start_pfn) {
 			size = end_pfn - start_pfn;
-			memmap_init_zone(size, nid, zone, start_pfn,
+			memmap_init_zone(size, nid, zone, start_pfn, range_end_pfn,
 					 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 		}
 

Signature · 1 line

2.17.2

The current memmap_init_zone() only handles memory region inside one zone, actually memmap_init() does the memmap init of one zone. So rename both of them accordingly.

And also rename the function parameter 'range_start_pfn' and local variable 'range_end_pfn' of memmap_init() to zone_start_pfn/zone_end_pfn.

Signed-off-by Baoquan He <[email protected]>

---
 arch/ia64/include/asm/pgtable.h |  2 +-
 arch/ia64/mm/init.c             |  6 +++---
 include/linux/mm.h              |  2 +-
 mm/memory_hotplug.c             |  2 +-
 mm/page_alloc.c                 | 24 ++++++++++++------------
 5 files changed, 18 insertions(+), 18 deletions(-)

arch/ia64/include/asm/pgtable.h +1 -1

index 779b6972aa84..dce2ff37df65 100644
@@ -520,7 +520,7 @@ extern struct page *zero_page_memmap_ptr;
 
 #  ifdef CONFIG_VIRTUAL_MEM_MAP
   /* arch mem_map init routine is needed due to holes in a virtual mem_map */
-    extern void memmap_init (unsigned long size, int nid, unsigned long zone,
+    extern void memmap_init_zone(unsigned long size, int nid, unsigned long zone,
 			     unsigned long start_pfn);
 #  endif /* CONFIG_VIRTUAL_MEM_MAP */
 # endif /* !__ASSEMBLY__ */

arch/ia64/mm/init.c +3 -3

index e76386a3479e..c8e68e92beb3 100644
@@ -535,18 +535,18 @@ virtual_memmap_init(u64 start, u64 end, void *arg)
 		    / sizeof(struct page));
 
 	if (map_start < map_end)
-		memmap_init_zone((unsigned long)(map_end - map_start),
+		memmap_init_range((unsigned long)(map_end - map_start),
 				 args->nid, args->zone, page_to_pfn(map_start), page_to_pfn(map_end),
 				 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 	return 0;
 }
 
 void __meminit
-memmap_init (unsigned long size, int nid, unsigned long zone,
+memmap_init_zone(unsigned long size, int nid, unsigned long zone,
 	     unsigned long start_pfn)
 {
 	if (!vmem_map) {
-		memmap_init_zone(size, nid, zone, start_pfn, start_pfn + size,
+		memmap_init_range(size, nid, zone, start_pfn, start_pfn + size,
 				 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 	} else {
 		struct page *start;

include/linux/mm.h +1 -1

index 92e06ea053f4..f72c138c2272 100644
@@ -2432,7 +2432,7 @@ extern int __meminit early_pfn_to_nid(unsigned long pfn);
 #endif
 
 extern void set_dma_reserve(unsigned long new_dma_reserve);
-extern void memmap_init_zone(unsigned long, int, unsigned long,
+extern void memmap_init_range(unsigned long, int, unsigned long,
 		unsigned long, unsigned long, enum meminit_context,
 		struct vmem_altmap *, int migratetype);
 extern void setup_per_zone_wmarks(void);

mm/memory_hotplug.c +1 -1

index f9d57b9be8c7..ddcb1cd24c60 100644
@@ -713,7 +713,7 @@ void __ref move_pfn_range_to_zone(struct zone *zone, unsigned long start_pfn,
 	 * expects the zone spans the pfn range. All the pages in the range
 	 * are reserved so nobody should be touching them so we should be safe
 	 */
-	memmap_init_zone(nr_pages, nid, zone_idx(zone), start_pfn, 0,
+	memmap_init_range(nr_pages, nid, zone_idx(zone), start_pfn, 0,
 			 MEMINIT_HOTPLUG, altmap, migratetype);
 
 	set_zone_contiguous(zone);

mm/page_alloc.c +12 -12

index 32645f2e7b96..4b46326099d9 100644
@@ -6117,7 +6117,7 @@ overlap_memmap_init(unsigned long zone, unsigned long *pfn)
  * (usually MIGRATE_MOVABLE). Besides setting the migratetype, no related
  * zone stats (e.g., nr_isolate_pageblock) are touched.
  */
-void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
+void __meminit memmap_init_range(unsigned long size, int nid, unsigned long zone,
 		unsigned long start_pfn, unsigned long zone_end_pfn,
 		enum meminit_context context,
 		struct vmem_altmap *altmap, int migratetype)

@@ -6292,24 +6292,24 @@ static inline u64 init_unavailable_range(unsigned long spfn, unsigned long epfn,
 }
 #endif
 
-void __init __weak memmap_init(unsigned long size, int nid,
+void __init __weak memmap_init_zone(unsigned long size, int nid,
 			       unsigned long zone,
-			       unsigned long range_start_pfn)
+			       unsigned long zone_start_pfn)
 {
 	unsigned long start_pfn, end_pfn, hole_start_pfn = 0;
-	unsigned long range_end_pfn = range_start_pfn + size;
+	unsigned long zone_end_pfn = zone_start_pfn + size;
 	u64 pgcnt = 0;
 	int i;
 
 	for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
-		start_pfn = clamp(start_pfn, range_start_pfn, range_end_pfn);
-		end_pfn = clamp(end_pfn, range_start_pfn, range_end_pfn);
-		hole_start_pfn = clamp(hole_start_pfn, range_start_pfn,
-				       range_end_pfn);
+		start_pfn = clamp(start_pfn, zone_start_pfn, zone_end_pfn);
+		end_pfn = clamp(end_pfn, zone_start_pfn, zone_end_pfn);
+		hole_start_pfn = clamp(hole_start_pfn, zone_start_pfn,
+				       zone_end_pfn);
 
 		if (end_pfn > start_pfn) {
 			size = end_pfn - start_pfn;
-			memmap_init_zone(size, nid, zone, start_pfn, range_end_pfn,
+			memmap_init_range(size, nid, zone, start_pfn, zone_end_pfn,
 					 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 		}
 

@@ -6326,8 +6326,8 @@ void __init __weak memmap_init(unsigned long size, int nid,
 	 * considered initialized. Make sure that memmap has a well defined
 	 * state.
 	 */
-	if (hole_start_pfn < range_end_pfn)
-		pgcnt += init_unavailable_range(hole_start_pfn, range_end_pfn,
+	if (hole_start_pfn < zone_end_pfn)
+		pgcnt += init_unavailable_range(hole_start_pfn, zone_end_pfn,
 						zone, nid);
 
 	if (pgcnt)

@@ -7039,7 +7039,7 @@ static void __init free_area_init_core(struct pglist_data *pgdat)
 		set_pageblock_order();
 		setup_usemap(pgdat, zone, zone_start_pfn, size);
 		init_currently_empty_zone(zone, zone_start_pfn, size);
-		memmap_init(size, nid, j, zone_start_pfn);
+		memmap_init_zone(size, nid, j, zone_start_pfn);
 	}
 }
 

Signature · 1 line

2.17.2

As David suggested, simply passing 'struct zone *zone' is enough. We can get all needed information from 'struct zone*' easily.

Suggested-by David Hildenbrand <[email protected]>

Signed-off-by Baoquan He <[email protected]>

---
 arch/ia64/include/asm/pgtable.h |  3 +--
 arch/ia64/mm/init.c             | 12 +++++++-----
 mm/page_alloc.c                 | 20 ++++++++++----------
 3 files changed, 18 insertions(+), 17 deletions(-)

arch/ia64/include/asm/pgtable.h +1 -2

index dce2ff37df65..2c81394a2430 100644
@@ -520,8 +520,7 @@ extern struct page *zero_page_memmap_ptr;
 
 #  ifdef CONFIG_VIRTUAL_MEM_MAP
   /* arch mem_map init routine is needed due to holes in a virtual mem_map */
-    extern void memmap_init_zone(unsigned long size, int nid, unsigned long zone,
-			     unsigned long start_pfn);
+    extern void memmap_init_zone(struct zone *zone);
 #  endif /* CONFIG_VIRTUAL_MEM_MAP */
 # endif /* !__ASSEMBLY__ */
 

arch/ia64/mm/init.c +7 -5

index c8e68e92beb3..ccbda1a74c95 100644
@@ -541,12 +541,14 @@ virtual_memmap_init(u64 start, u64 end, void *arg)
 	return 0;
 }
 
-void __meminit
-memmap_init_zone(unsigned long size, int nid, unsigned long zone,
-	     unsigned long start_pfn)
+void __meminit memmap_init_zone(struct zone *zone)
 {
+	unsigned long size = zone->spanned_pages;
+	int nid = zone_to_nid(zone), zone_id = zone_idx(zone);
+	unsigned long start_pfn = zone->zone_start_pfn;
+
 	if (!vmem_map) {
-		memmap_init_range(size, nid, zone, start_pfn, start_pfn + size,
+		memmap_init_range(size, nid, zone_id, start_pfn, start_pfn + size,
 				 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 	} else {
 		struct page *start;

@@ -556,7 +558,7 @@ memmap_init_zone(unsigned long size, int nid, unsigned long zone,
 		args.start = start;
 		args.end = start + size;
 		args.nid = nid;
-		args.zone = zone;
+		args.zone = zone_id;
 
 		efi_memmap_walk(virtual_memmap_init, &args);
 	}

mm/page_alloc.c +10 -10

index 4b46326099d9..7a6626351ed7 100644
@@ -6292,16 +6292,16 @@ static inline u64 init_unavailable_range(unsigned long spfn, unsigned long epfn,
 }
 #endif
 
-void __init __weak memmap_init_zone(unsigned long size, int nid,
-			       unsigned long zone,
-			       unsigned long zone_start_pfn)
+void __init __weak memmap_init_zone(struct zone *zone)
 {
 	unsigned long start_pfn, end_pfn, hole_start_pfn = 0;
-	unsigned long zone_end_pfn = zone_start_pfn + size;
+	int i, nid = zone_to_nid(zone), zone_id = zone_idx(zone);
+	unsigned long zone_start_pfn = zone->zone_start_pfn;
+	unsigned long zone_end_pfn = zone_start_pfn + zone->spanned_pages;
 	u64 pgcnt = 0;
-	int i;
 
 	for_each_mem_pfn_range(i, nid, &start_pfn, &end_pfn, NULL) {
+		unsigned long size;
 		start_pfn = clamp(start_pfn, zone_start_pfn, zone_end_pfn);
 		end_pfn = clamp(end_pfn, zone_start_pfn, zone_end_pfn);
 		hole_start_pfn = clamp(hole_start_pfn, zone_start_pfn,

@@ -6309,13 +6309,13 @@ void __init __weak memmap_init_zone(unsigned long size, int nid,
 
 		if (end_pfn > start_pfn) {
 			size = end_pfn - start_pfn;
-			memmap_init_range(size, nid, zone, start_pfn, zone_end_pfn,
+			memmap_init_range(size, nid, zone_id, start_pfn, zone_end_pfn,
 					 MEMINIT_EARLY, NULL, MIGRATE_MOVABLE);
 		}
 
 		if (hole_start_pfn < start_pfn)
 			pgcnt += init_unavailable_range(hole_start_pfn,
-							start_pfn, zone, nid);
+							start_pfn, zone_id, nid);
 		hole_start_pfn = end_pfn;
 	}
 

@@ -6328,11 +6328,11 @@ void __init __weak memmap_init_zone(unsigned long size, int nid,
 	 */
 	if (hole_start_pfn < zone_end_pfn)
 		pgcnt += init_unavailable_range(hole_start_pfn, zone_end_pfn,
-						zone, nid);
+						zone_id, nid);
 
 	if (pgcnt)
 		pr_info("%s: Zeroed struct page in unavailable ranges: %lld\n",
-			zone_names[zone], pgcnt);
+			zone_names[zone_id], pgcnt);
 }
 
 static int zone_batchsize(struct zone *zone)

@@ -7039,7 +7039,7 @@ static void __init free_area_init_core(struct pglist_data *pgdat)
 		set_pageblock_order();
 		setup_usemap(pgdat, zone, zone_start_pfn, size);
 		init_currently_empty_zone(zone, zone_start_pfn, size);
-		memmap_init_zone(size, nid, j, zone_start_pfn);
+		memmap_init_zone(zone);
 	}
 }
 

Signature · 1 line

2.17.2

Parameter 'zone' has got needed information, let's remove other unnecessary parameters.

Signed-off-by Baoquan He <[email protected]>

---
 mm/page_alloc.c | 17 +++++++----------
 1 file changed, 7 insertions(+), 10 deletions(-)

mm/page_alloc.c +7 -10

index 7a6626351ed7..7f0a917ab858 100644
@@ -6824,25 +6824,22 @@ static unsigned long __init usemap_size(unsigned long zone_start_pfn, unsigned l
 	return usemapsize / 8;
 }
 
-static void __ref setup_usemap(struct pglist_data *pgdat,
-				struct zone *zone,
-				unsigned long zone_start_pfn,
-				unsigned long zonesize)
+static void __ref setup_usemap(struct zone *zone)
 {
-	unsigned long usemapsize = usemap_size(zone_start_pfn, zonesize);
+	unsigned long usemapsize = usemap_size(zone->zone_start_pfn,
+					       zone->spanned_pages);
 	zone->pageblock_flags = NULL;
 	if (usemapsize) {
 		zone->pageblock_flags =
 			memblock_alloc_node(usemapsize, SMP_CACHE_BYTES,
-					    pgdat->node_id);
+					    zone_to_nid(zone));
 		if (!zone->pageblock_flags)
 			panic("Failed to allocate %ld bytes for zone %s pageblock flags on node %d\n",
-			      usemapsize, zone->name, pgdat->node_id);
+			      usemapsize, zone->name, zone_to_nid(zone));
 	}
 }
 #else
-static inline void setup_usemap(struct pglist_data *pgdat, struct zone *zone,
-				unsigned long zone_start_pfn, unsigned long zonesize) {}
+static inline void setup_usemap(struct zone *zone) {}
 #endif /* CONFIG_SPARSEMEM */
 
 #ifdef CONFIG_HUGETLB_PAGE_SIZE_VARIABLE

@@ -7037,7 +7034,7 @@ static void __init free_area_init_core(struct pglist_data *pgdat)
 			continue;
 
 		set_pageblock_order();
-		setup_usemap(pgdat, zone, zone_start_pfn, size);
+		setup_usemap(zone);
 		init_currently_empty_zone(zone, zone_start_pfn, size);
 		memmap_init_zone(zone);
 	}

Signature · 1 line

2.17.2

Local variable 'zone_start_pfn' is not needed since there's only one call site in free_area_init_core(). Let's remove it and pass zone->zone_start_pfn directly to init_currently_empty_zone().

Signed-off-by Baoquan He <[email protected]>

---
 mm/page_alloc.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

mm/page_alloc.c +1 -2

index 7f0a917ab858..189a86253c93 100644
@@ -6986,7 +6986,6 @@ static void __init free_area_init_core(struct pglist_data *pgdat)
 	for (j = 0; j < MAX_NR_ZONES; j++) {
 		struct zone *zone = pgdat->node_zones + j;
 		unsigned long size, freesize, memmap_pages;
-		unsigned long zone_start_pfn = zone->zone_start_pfn;
 
 		size = zone->spanned_pages;
 		freesize = zone->present_pages;

@@ -7035,7 +7034,7 @@ static void __init free_area_init_core(struct pglist_data *pgdat)
 
 		set_pageblock_order();
 		setup_usemap(zone);
-		init_currently_empty_zone(zone, zone_start_pfn, size);
+		init_currently_empty_zone(zone, zone->zone_start_pfn, size);
 		memmap_init_zone(zone);
 	}
 }

Signature · 1 line

2.17.2

VMware reported the performance regression during memmap_init() invocation. And they bisected to commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN") causing it.

https://lore.kernel.org/linux-mm/DM6PR05MB52921FF90FA01CC337DD23A1A4080@DM6PR05MB5292.namprd05.prod.outlook.com/

After investigation, it's caused by incorrect memmap init defer handling in memmap_init_zone() after commit 73a6e474cb376. The current memmap_init_zone() only handle one memory region of one zone, while memmap_init() iterates over all its memory regions and pass them one by one into memmap_init_zone() to handle.

So in this patchset, patch 1/5 fixes the bug observed by VMware. Patch 2~5/5 clean up codes.
accordingly.

Quoted 12 lines · On Sun, 20 Dec 2020 16:27:49 +0800 Baoquan He <[email protected]> wrote:

VMware reported the performance regression during memmap_init() invocation. And they bisected to commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN") causing it.

https://lore.kernel.org/linux-mm/DM6PR05MB52921FF90FA01CC337DD23A1A4080@DM6PR05MB5292.namprd05.prod.outlook.com/

After investigation, it's caused by incorrect memmap init defer handling in memmap_init_zone() after commit 73a6e474cb376. The current memmap_init_zone() only handle one memory region of one zone, while memmap_init() iterates over all its memory regions and pass them one by one into memmap_init_zone() to handle.

Quoted 3 lines

So in this patchset, patch 1/5 fixes the bug observed by VMware. Patch 2~5/5 clean up codes.
accordingly.

This series doesn't apply well to current mainline (plus, perhaps, material which I sent to Linus today).

So please check all that against mainline in a day or so, refresh, retest and resend.

Please separate the fix for the performance regression (1/5) into a single standalone patch, ready for -stable backporting. And then a separate 4-patch series with the cleanups for a 5.11 merge.

Quoted 21 lines · Andrew Morton

Quoted 12 lines · On Sun, 20 Dec 2020 16:27:49 +0800 Baoquan He <[email protected]> wrote:

VMware reported the performance regression during memmap_init() invocation. And they bisected to commit 73a6e474cb376 ("mm: memmap_init: iterate over memblock regions rather that check each PFN") causing it.

https://lore.kernel.org/linux-mm/DM6PR05MB52921FF90FA01CC337DD23A1A4080@DM6PR05MB5292.namprd05.prod.outlook.com/

After investigation, it's caused by incorrect memmap init defer handling in memmap_init_zone() after commit 73a6e474cb376. The current memmap_init_zone() only handle one memory region of one zone, while memmap_init() iterates over all its memory regions and pass them one by one into memmap_init_zone() to handle.

Quoted 3 lines

So in this patchset, patch 1/5 fixes the bug observed by VMware. Patch 2~5/5 clean up codes.
accordingly.

This series doesn't apply well to current mainline (plus, perhaps, material which I sent to Linus today).

Quoted 6 lines

So please check all that against mainline in a day or so, refresh, retest and resend.

Please separate the fix for the performance regression (1/5) into a single standalone patch, ready for -stable backporting. And then a separate 4-patch series with the cleanups for a 5.11 merge.

Sure, doing now.

By the way, when sending patches to linux-mm ML, which branch should I rebase them on? I usually take your akpm/master as base, thought this will make your patch picking easier. Seems my understanding is not true, akpm/master is changed very soon, we should always base patch on linus's master branch, whether patch is sending to linux-mm or not, right?