[PATCH v2 0/5] Cleanups and fixup for page_alloc

[PATCH v2 0/5] Cleanups and fixup for page_alloc

[Miaohe Lin]

Hi all,
This series contains cleanups to remove meaningless VM_BUG_ON(), use
helpers to simplify the code and remove obsolete comment. Also we avoid
allocating highmem pages via alloc_pages_exact[_nid]. More details can
be found in the respective changelogs. Thanks!

v1->v2:
  collect Reviewed-by and Acked-by tag
  preserve valid comments for free_pcppages_bulk
  send patch "mm/page_alloc.c: avoid accessing uninitialized pcp page
migratetype" separately.
  s/__GFP_COMP/(__GFP_COMP|__GFP_HIGHMEM)/g throughout both
alloc_pages_exact() and alloc_pages_exact_nid()

Miaohe Lin (5):
  mm/page_alloc.c: remove meaningless VM_BUG_ON() in pindex_to_order()
  mm/page_alloc.c: simplify the code by using macro K()
  mm/page_alloc.c: fix obsolete comment in free_pcppages_bulk()
  mm/page_alloc.c: use helper function zone_spans_pfn()
  mm/page_alloc.c: avoid allocating highmem pages via
    alloc_pages_exact[_nid]

 mm/page_alloc.c | 34 ++++++++++++----------------------
 1 file changed, 12 insertions(+), 22 deletions(-)

-- 
2.23.0

[PATCH v2 1/5] mm/page_alloc.c: remove meaningless VM_BUG_ON() in pindex_to_order()

[Miaohe Lin]

It's meaningless to VM_BUG_ON() order != pageblock_order just after
setting order to pageblock_order. Remove it.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: David Hildenbrand <david@redhat.com>
---
 mm/page_alloc.c | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 91edb930b8ab..dbb3338d9287 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -677,10 +677,8 @@ static inline int pindex_to_order(unsigned int pindex)
 	int order = pindex / MIGRATE_PCPTYPES;
 
 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
-	if (order > PAGE_ALLOC_COSTLY_ORDER) {
+	if (order > PAGE_ALLOC_COSTLY_ORDER)
 		order = pageblock_order;
-		VM_BUG_ON(order != pageblock_order);
-	}
 #else
 	VM_BUG_ON(order > PAGE_ALLOC_COSTLY_ORDER);
 #endif
-- 
2.23.0

[PATCH v2 2/5] mm/page_alloc.c: simplify the code by using macro K()

[Miaohe Lin]

Use helper macro K() to convert the pages to the corresponding size.
Minor readability improvement.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: David Hildenbrand <david@redhat.com>
---
 mm/page_alloc.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index dbb3338d9287..d3983244f56f 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -8134,8 +8134,7 @@ unsigned long free_reserved_area(void *start, void *end, int poison, const char
 	}
 
 	if (pages && s)
-		pr_info("Freeing %s memory: %ldK\n",
-			s, pages << (PAGE_SHIFT - 10));
+		pr_info("Freeing %s memory: %ldK\n", s, K(pages));
 
 	return pages;
 }
@@ -8180,14 +8179,13 @@ void __init mem_init_print_info(void)
 		", %luK highmem"
 #endif
 		")\n",
-		nr_free_pages() << (PAGE_SHIFT - 10),
-		physpages << (PAGE_SHIFT - 10),
+		K(nr_free_pages()), K(physpages),
 		codesize >> 10, datasize >> 10, rosize >> 10,
 		(init_data_size + init_code_size) >> 10, bss_size >> 10,
-		(physpages - totalram_pages() - totalcma_pages) << (PAGE_SHIFT - 10),
-		totalcma_pages << (PAGE_SHIFT - 10)
+		K(physpages - totalram_pages() - totalcma_pages),
+		K(totalcma_pages)
 #ifdef	CONFIG_HIGHMEM
-		, totalhigh_pages() << (PAGE_SHIFT - 10)
+		, K(totalhigh_pages())
 #endif
 		);
 }
-- 
2.23.0

[PATCH v2 3/5] mm/page_alloc.c: fix obsolete comment in free_pcppages_bulk()

[Miaohe Lin]

The second two paragraphs about "all pages pinned" and pages_scanned is
obsolete. And There are PAGE_ALLOC_COSTLY_ORDER + 1 + NR_PCP_THP orders
in pcp. So the same order assumption is not held now.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
---
 mm/page_alloc.c | 8 +-------
 1 file changed, 1 insertion(+), 7 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index d3983244f56f..f7fcbac2d091 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1426,14 +1426,8 @@ static inline void prefetch_buddy(struct page *page)
 
 /*
  * Frees a number of pages from the PCP lists
- * Assumes all pages on list are in same zone, and of same order.
+ * Assumes all pages on list are in same zone.
  * count is the number of pages to free.
- *
- * If the zone was previously in an "all pages pinned" state then look to
- * see if this freeing clears that state.
- *
- * And clear the zone's pages_scanned counter, to hold off the "all pages are
- * pinned" detection logic.
  */
 static void free_pcppages_bulk(struct zone *zone, int count,
 					struct per_cpu_pages *pcp)
-- 
2.23.0

Re: [PATCH v2 3/5] mm/page_alloc.c: fix obsolete comment in free_pcppages_bulk()

[Mel Gorman]

On Thu, Sep 02, 2021 at 08:12:40PM +0800, Miaohe Lin wrote:
> The second two paragraphs about "all pages pinned" and pages_scanned is
> obsolete. And There are PAGE_ALLOC_COSTLY_ORDER + 1 + NR_PCP_THP orders
> in pcp. So the same order assumption is not held now.
> 
> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>

Acked-by: Mel Gorman <mgorman@techsingularity.net>

-- 
Mel Gorman
SUSE Labs

[PATCH v2 4/5] mm/page_alloc.c: use helper function zone_spans_pfn()

[Miaohe Lin]

Use helper function zone_spans_pfn() to check whether pfn is within a
zone to simplify the code slightly.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: David Hildenbrand <david@redhat.com>
---
 mm/page_alloc.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index f7fcbac2d091..9c09dcb24149 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -1581,7 +1581,7 @@ static void __meminit init_reserved_page(unsigned long pfn)
 	for (zid = 0; zid < MAX_NR_ZONES; zid++) {
 		struct zone *zone = &pgdat->node_zones[zid];
 
-		if (pfn >= zone->zone_start_pfn && pfn < zone_end_pfn(zone))
+		if (zone_spans_pfn(zone, pfn))
 			break;
 	}
 	__init_single_page(pfn_to_page(pfn), pfn, zid, nid);
-- 
2.23.0

[PATCH v2 5/5] mm/page_alloc.c: avoid allocating highmem pages via alloc_pages_exact[_nid]

[Miaohe Lin]

Don't use with __GFP_HIGHMEM because page_address() cannot represent
highmem pages without kmap(). Newly allocated pages would leak as
page_address() will return NULL for highmem pages here. But It works
now because the callers do not specify __GFP_HIGHMEM now.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
---
 mm/page_alloc.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 9c09dcb24149..e1d0e27d005a 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -5614,8 +5614,8 @@ void *alloc_pages_exact(size_t size, gfp_t gfp_mask)
 	unsigned int order = get_order(size);
 	unsigned long addr;
 
-	if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
-		gfp_mask &= ~__GFP_COMP;
+	if (WARN_ON_ONCE(gfp_mask & (__GFP_COMP | __GFP_HIGHMEM)))
+		gfp_mask &= ~(__GFP_COMP | __GFP_HIGHMEM);
 
 	addr = __get_free_pages(gfp_mask, order);
 	return make_alloc_exact(addr, order, size);
@@ -5639,8 +5639,8 @@ void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask)
 	unsigned int order = get_order(size);
 	struct page *p;
 
-	if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
-		gfp_mask &= ~__GFP_COMP;
+	if (WARN_ON_ONCE(gfp_mask & (__GFP_COMP | __GFP_HIGHMEM)))
+		gfp_mask &= ~(__GFP_COMP | __GFP_HIGHMEM);
 
 	p = alloc_pages_node(nid, gfp_mask, order);
 	if (!p)
-- 
2.23.0

Re: [PATCH v2 5/5] mm/page_alloc.c: avoid allocating highmem pages via alloc_pages_exact[_nid]

[David Hildenbrand]

On 02.09.21 14:12, Miaohe Lin wrote:
> Don't use with __GFP_HIGHMEM because page_address() cannot represent
> highmem pages without kmap(). Newly allocated pages would leak as
> page_address() will return NULL for highmem pages here. But It works
> now because the callers do not specify __GFP_HIGHMEM now.
> 
> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
> ---
>   mm/page_alloc.c | 8 ++++----
>   1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index 9c09dcb24149..e1d0e27d005a 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5614,8 +5614,8 @@ void *alloc_pages_exact(size_t size, gfp_t gfp_mask)
>   	unsigned int order = get_order(size);
>   	unsigned long addr;
>   
> -	if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
> -		gfp_mask &= ~__GFP_COMP;
> +	if (WARN_ON_ONCE(gfp_mask & (__GFP_COMP | __GFP_HIGHMEM)))
> +		gfp_mask &= ~(__GFP_COMP | __GFP_HIGHMEM);
>   
>   	addr = __get_free_pages(gfp_mask, order);
>   	return make_alloc_exact(addr, order, size);
> @@ -5639,8 +5639,8 @@ void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask)
>   	unsigned int order = get_order(size);
>   	struct page *p;
>   
> -	if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
> -		gfp_mask &= ~__GFP_COMP;
> +	if (WARN_ON_ONCE(gfp_mask & (__GFP_COMP | __GFP_HIGHMEM)))
> +		gfp_mask &= ~(__GFP_COMP | __GFP_HIGHMEM);
>   
>   	p = alloc_pages_node(nid, gfp_mask, order);
>   	if (!p)
> 

Ideally we would convert this WARN_ON_ONCE() to pr_warn_once(), but I 
guess this really never ever happens on a production system and would 
get caught early while testing.

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 
Thanks,

David / dhildenb

Re: [PATCH v2 5/5] mm/page_alloc.c: avoid allocating highmem pages via alloc_pages_exact[_nid]

[Miaohe Lin]

On 2021/9/2 20:16, David Hildenbrand wrote:
> On 02.09.21 14:12, Miaohe Lin wrote:
>> Don't use with __GFP_HIGHMEM because page_address() cannot represent
>> highmem pages without kmap(). Newly allocated pages would leak as
>> page_address() will return NULL for highmem pages here. But It works
>> now because the callers do not specify __GFP_HIGHMEM now.
>>
>> Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
>> ---
>>   mm/page_alloc.c | 8 ++++----
>>   1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index 9c09dcb24149..e1d0e27d005a 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -5614,8 +5614,8 @@ void *alloc_pages_exact(size_t size, gfp_t gfp_mask)
>>       unsigned int order = get_order(size);
>>       unsigned long addr;
>>   -    if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
>> -        gfp_mask &= ~__GFP_COMP;
>> +    if (WARN_ON_ONCE(gfp_mask & (__GFP_COMP | __GFP_HIGHMEM)))
>> +        gfp_mask &= ~(__GFP_COMP | __GFP_HIGHMEM);
>>         addr = __get_free_pages(gfp_mask, order);
>>       return make_alloc_exact(addr, order, size);
>> @@ -5639,8 +5639,8 @@ void * __meminit alloc_pages_exact_nid(int nid, size_t size, gfp_t gfp_mask)
>>       unsigned int order = get_order(size);
>>       struct page *p;
>>   -    if (WARN_ON_ONCE(gfp_mask & __GFP_COMP))
>> -        gfp_mask &= ~__GFP_COMP;
>> +    if (WARN_ON_ONCE(gfp_mask & (__GFP_COMP | __GFP_HIGHMEM)))
>> +        gfp_mask &= ~(__GFP_COMP | __GFP_HIGHMEM);
>>         p = alloc_pages_node(nid, gfp_mask, order);
>>       if (!p)
>>
> 
> Ideally we would convert this WARN_ON_ONCE() to pr_warn_once(), but I guess this really never ever happens on a production system and would get caught early while testing.
> 

This patch is mainly intended to make it clear that we simply can't call
alloc_pages_exact[_nid] with GFP_HIGHMEM. So this warning would never ever happens.

> Reviewed-by: David Hildenbrand <david@redhat.com>
> 

Many thanks for your review and reply.