[2/2] mm: compaction: skip the memory hole rapidly when isolating free pages

Message ID d2ba7e41ee566309b594311207ffca736375fc16.1688715750.git.baolin.wang@linux.alibaba.com
State New
Headers
Series [1/2] mm: compaction: use the correct type of list for free pages |

Commit Message

Baolin Wang July 7, 2023, 8:51 a.m. UTC
  On my machine with below memory layout, and I can see it will take more
time to skip the larger memory hole (range: 0x100000000 - 0x1800000000)
when isolating free pages. So adding a new helper to skip the memory
hole rapidly, which can reduce the time consumed from about 70us to less
than 1us.

[    0.000000] Zone ranges:
[    0.000000]   DMA      [mem 0x0000000040000000-0x00000000ffffffff]
[    0.000000]   DMA32    empty
[    0.000000]   Normal   [mem 0x0000000100000000-0x0000001fa7ffffff]
[    0.000000] Movable zone start for each node
[    0.000000] Early memory node ranges
[    0.000000]   node   0: [mem 0x0000000040000000-0x0000000fffffffff]
[    0.000000]   node   0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
[    0.000000]   node   0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
[    0.000000]   node   0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
[    0.000000]   node   0: [mem 0x0000001fa4030000-0x0000001fa40effff]
[    0.000000]   node   0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
[    0.000000]   node   0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
[    0.000000]   node   0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
[    0.000000]   node   0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
[    0.000000]   node   0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]

Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
---
 mm/compaction.c | 30 +++++++++++++++++++++++++++++-
 1 file changed, 29 insertions(+), 1 deletion(-)
  

Comments

David Hildenbrand July 7, 2023, 11:56 a.m. UTC | #1
On 07.07.23 10:51, Baolin Wang wrote:
> On my machine with below memory layout, and I can see it will take more
> time to skip the larger memory hole (range: 0x100000000 - 0x1800000000)
> when isolating free pages. So adding a new helper to skip the memory
> hole rapidly, which can reduce the time consumed from about 70us to less
> than 1us.

Can you clarify how this relates to the previous commit and mention that 
commit?

> 
> [    0.000000] Zone ranges:
> [    0.000000]   DMA      [mem 0x0000000040000000-0x00000000ffffffff]
> [    0.000000]   DMA32    empty
> [    0.000000]   Normal   [mem 0x0000000100000000-0x0000001fa7ffffff]
> [    0.000000] Movable zone start for each node
> [    0.000000] Early memory node ranges
> [    0.000000]   node   0: [mem 0x0000000040000000-0x0000000fffffffff]
> [    0.000000]   node   0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
> [    0.000000]   node   0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
> [    0.000000]   node   0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
> [    0.000000]   node   0: [mem 0x0000001fa4030000-0x0000001fa40effff]
> [    0.000000]   node   0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
> [    0.000000]   node   0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]
> 
> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> ---
>   mm/compaction.c | 30 +++++++++++++++++++++++++++++-
>   1 file changed, 29 insertions(+), 1 deletion(-)
> 
> diff --git a/mm/compaction.c b/mm/compaction.c
> index 43358efdbdc2..9641e2131901 100644
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -249,11 +249,31 @@ static unsigned long skip_offline_sections(unsigned long start_pfn)
>   
>   	return 0;
>   }
> +
> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
> +{
> +	unsigned long start_nr = pfn_to_section_nr(start_pfn);
> +
> +	if (!start_nr || online_section_nr(start_nr))
> +		return 0;
> +
> +	while (start_nr-- > 0) {
> +		if (online_section_nr(start_nr))
> +			return section_nr_to_pfn(start_nr) + PAGES_PER_SECTION - 1;
> +	}
> +
> +	return 0;
> +}
>   #else
>   static unsigned long skip_offline_sections(unsigned long start_pfn)
>   {
>   	return 0;
>   }
> +
> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
> +{
> +	return 0;
> +}
>   #endif
>   
>   /*
> @@ -1668,8 +1688,16 @@ static void isolate_freepages(struct compact_control *cc)
>   
>   		page = pageblock_pfn_to_page(block_start_pfn, block_end_pfn,
>   									zone);
> -		if (!page)
> +		if (!page) {
> +			unsigned long next_pfn;
> +
> +			next_pfn = skip_offline_sections_reverse(block_start_pfn);
> +			if (next_pfn)
> +				block_start_pfn = max(pageblock_start_pfn(next_pfn),
> +						      low_pfn);
> +
>   			continue;
> +		}
>   
>   		/* Check the block is suitable for migration */
>   		if (!suitable_migration_target(cc, page))


LGTM

Acked-by: David Hildenbrand <david@redhat.com>
  
Huang, Ying July 10, 2023, 6:11 a.m. UTC | #2
Baolin Wang <baolin.wang@linux.alibaba.com> writes:

> On my machine with below memory layout, and I can see it will take more
> time to skip the larger memory hole (range: 0x100000000 - 0x1800000000)
> when isolating free pages. So adding a new helper to skip the memory
> hole rapidly, which can reduce the time consumed from about 70us to less
> than 1us.
>
> [    0.000000] Zone ranges:
> [    0.000000]   DMA      [mem 0x0000000040000000-0x00000000ffffffff]
> [    0.000000]   DMA32    empty
> [    0.000000]   Normal   [mem 0x0000000100000000-0x0000001fa7ffffff]

The memory hole is at the beginning of zone NORMAL?  If so, should zone
NORMAL start at 0x1800000000?  And, the free pages will not be scanned
there?  Or my understanding were wrong?

Best Regards,
Huang, Ying

> [    0.000000] Movable zone start for each node
> [    0.000000] Early memory node ranges
> [    0.000000]   node   0: [mem 0x0000000040000000-0x0000000fffffffff]
> [    0.000000]   node   0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
> [    0.000000]   node   0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
> [    0.000000]   node   0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
> [    0.000000]   node   0: [mem 0x0000001fa4030000-0x0000001fa40effff]
> [    0.000000]   node   0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
> [    0.000000]   node   0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]
>
> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
> ---
>  mm/compaction.c | 30 +++++++++++++++++++++++++++++-
>  1 file changed, 29 insertions(+), 1 deletion(-)
>
> diff --git a/mm/compaction.c b/mm/compaction.c
> index 43358efdbdc2..9641e2131901 100644
> --- a/mm/compaction.c
> +++ b/mm/compaction.c
> @@ -249,11 +249,31 @@ static unsigned long skip_offline_sections(unsigned long start_pfn)
>  
>  	return 0;
>  }
> +
> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
> +{
> +	unsigned long start_nr = pfn_to_section_nr(start_pfn);
> +
> +	if (!start_nr || online_section_nr(start_nr))
> +		return 0;
> +
> +	while (start_nr-- > 0) {
> +		if (online_section_nr(start_nr))
> +			return section_nr_to_pfn(start_nr) + PAGES_PER_SECTION - 1;
> +	}
> +
> +	return 0;
> +}
>  #else
>  static unsigned long skip_offline_sections(unsigned long start_pfn)
>  {
>  	return 0;
>  }
> +
> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
> +{
> +	return 0;
> +}
>  #endif
>  
>  /*
> @@ -1668,8 +1688,16 @@ static void isolate_freepages(struct compact_control *cc)
>  
>  		page = pageblock_pfn_to_page(block_start_pfn, block_end_pfn,
>  									zone);
> -		if (!page)
> +		if (!page) {
> +			unsigned long next_pfn;
> +
> +			next_pfn = skip_offline_sections_reverse(block_start_pfn);
> +			if (next_pfn)
> +				block_start_pfn = max(pageblock_start_pfn(next_pfn),
> +						      low_pfn);
> +
>  			continue;
> +		}
>  
>  		/* Check the block is suitable for migration */
>  		if (!suitable_migration_target(cc, page))
  
Baolin Wang July 10, 2023, 9:26 a.m. UTC | #3
On 7/10/2023 2:11 PM, Huang, Ying wrote:
> Baolin Wang <baolin.wang@linux.alibaba.com> writes:
> 
>> On my machine with below memory layout, and I can see it will take more
>> time to skip the larger memory hole (range: 0x100000000 - 0x1800000000)
>> when isolating free pages. So adding a new helper to skip the memory
>> hole rapidly, which can reduce the time consumed from about 70us to less
>> than 1us.
>>
>> [    0.000000] Zone ranges:
>> [    0.000000]   DMA      [mem 0x0000000040000000-0x00000000ffffffff]
>> [    0.000000]   DMA32    empty
>> [    0.000000]   Normal   [mem 0x0000000100000000-0x0000001fa7ffffff]
> 
> The memory hole is at the beginning of zone NORMAL?  If so, should zone

No, the memory hole range is 0x1000000000 - 0x1800000000, and the normal 
zone is start from 0x100000000.

I'm sorry I made a typo in the commit message, which confuses you. The 
memory hole range should be: 0x1000000000 - 0x1800000000. I updated the 
commit message to the following and addressed David's comment:

"
Just like commit 9721fd82351d ("mm: compaction: skip memory hole rapidly
when isolating migratable pages"), I can see it will also take more
time to skip the larger memory hole (range: 0x1000000000 - 0x1800000000)
when isolating free pages on my machine with below memory layout. So 
like commit 9721fd82351d, adding a new helper to skip the memory hole 
rapidly, which can reduce the time consumed from about 70us to less than 
1us.

[    0.000000] Zone ranges:
[    0.000000]   DMA      [mem 0x0000000040000000-0x00000000ffffffff]
[    0.000000]   DMA32    empty
[    0.000000]   Normal   [mem 0x0000000100000000-0x0000001fa7ffffff]
[    0.000000] Movable zone start for each node
[    0.000000] Early memory node ranges
[    0.000000]   node   0: [mem 0x0000000040000000-0x0000000fffffffff]
[    0.000000]   node   0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
[    0.000000]   node   0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
[    0.000000]   node   0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
[    0.000000]   node   0: [mem 0x0000001fa4030000-0x0000001fa40effff]
[    0.000000]   node   0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
[    0.000000]   node   0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
[    0.000000]   node   0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
[    0.000000]   node   0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
[    0.000000]   node   0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]
"

> NORMAL start at 0x1800000000?  And, the free pages will not be scanned
> there?  Or my understanding were wrong. >
>> [    0.000000] Movable zone start for each node
>> [    0.000000] Early memory node ranges
>> [    0.000000]   node   0: [mem 0x0000000040000000-0x0000000fffffffff]
>> [    0.000000]   node   0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
>> [    0.000000]   node   0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
>> [    0.000000]   node   0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
>> [    0.000000]   node   0: [mem 0x0000001fa4030000-0x0000001fa40effff]
>> [    0.000000]   node   0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
>> [    0.000000]   node   0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
>> [    0.000000]   node   0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
>> [    0.000000]   node   0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
>> [    0.000000]   node   0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]
>>
>> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
>> ---
>>   mm/compaction.c | 30 +++++++++++++++++++++++++++++-
>>   1 file changed, 29 insertions(+), 1 deletion(-)
>>
>> diff --git a/mm/compaction.c b/mm/compaction.c
>> index 43358efdbdc2..9641e2131901 100644
>> --- a/mm/compaction.c
>> +++ b/mm/compaction.c
>> @@ -249,11 +249,31 @@ static unsigned long skip_offline_sections(unsigned long start_pfn)
>>   
>>   	return 0;
>>   }
>> +
>> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
>> +{
>> +	unsigned long start_nr = pfn_to_section_nr(start_pfn);
>> +
>> +	if (!start_nr || online_section_nr(start_nr))
>> +		return 0;
>> +
>> +	while (start_nr-- > 0) {
>> +		if (online_section_nr(start_nr))
>> +			return section_nr_to_pfn(start_nr) + PAGES_PER_SECTION - 1;
>> +	}
>> +
>> +	return 0;
>> +}
>>   #else
>>   static unsigned long skip_offline_sections(unsigned long start_pfn)
>>   {
>>   	return 0;
>>   }
>> +
>> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
>> +{
>> +	return 0;
>> +}
>>   #endif
>>   
>>   /*
>> @@ -1668,8 +1688,16 @@ static void isolate_freepages(struct compact_control *cc)
>>   
>>   		page = pageblock_pfn_to_page(block_start_pfn, block_end_pfn,
>>   									zone);
>> -		if (!page)
>> +		if (!page) {
>> +			unsigned long next_pfn;
>> +
>> +			next_pfn = skip_offline_sections_reverse(block_start_pfn);
>> +			if (next_pfn)
>> +				block_start_pfn = max(pageblock_start_pfn(next_pfn),
>> +						      low_pfn);
>> +
>>   			continue;
>> +		}
>>   
>>   		/* Check the block is suitable for migration */
>>   		if (!suitable_migration_target(cc, page))
  
Huang, Ying July 11, 2023, 12:37 a.m. UTC | #4
Baolin Wang <baolin.wang@linux.alibaba.com> writes:

> On 7/10/2023 2:11 PM, Huang, Ying wrote:
>> Baolin Wang <baolin.wang@linux.alibaba.com> writes:
>> 
>>> On my machine with below memory layout, and I can see it will take more
>>> time to skip the larger memory hole (range: 0x100000000 - 0x1800000000)
>>> when isolating free pages. So adding a new helper to skip the memory
>>> hole rapidly, which can reduce the time consumed from about 70us to less
>>> than 1us.
>>>
>>> [    0.000000] Zone ranges:
>>> [    0.000000]   DMA      [mem 0x0000000040000000-0x00000000ffffffff]
>>> [    0.000000]   DMA32    empty
>>> [    0.000000]   Normal   [mem 0x0000000100000000-0x0000001fa7ffffff]
>> The memory hole is at the beginning of zone NORMAL?  If so, should
>> zone
>
> No, the memory hole range is 0x1000000000 - 0x1800000000, and the
> normal zone is start from 0x100000000.
>
> I'm sorry I made a typo in the commit message, which confuses you. The
> memory hole range should be: 0x1000000000 - 0x1800000000. I updated
> the commit message to the following and addressed David's comment:

Got it!  Thanks for explanation!

> "
> Just like commit 9721fd82351d ("mm: compaction: skip memory hole rapidly
> when isolating migratable pages"), I can see it will also take more
> time to skip the larger memory hole (range: 0x1000000000 - 0x1800000000)
> when isolating free pages on my machine with below memory layout. So
> like commit 9721fd82351d, adding a new helper to skip the memory hole
> rapidly, which can reduce the time consumed from about 70us to less
> than 1us.

LGTM.

Reviewed-by: "Huang, Ying" <ying.huang@intel.com>

> [    0.000000] Zone ranges:
> [    0.000000]   DMA      [mem 0x0000000040000000-0x00000000ffffffff]
> [    0.000000]   DMA32    empty
> [    0.000000]   Normal   [mem 0x0000000100000000-0x0000001fa7ffffff]
> [    0.000000] Movable zone start for each node
> [    0.000000] Early memory node ranges
> [    0.000000]   node   0: [mem 0x0000000040000000-0x0000000fffffffff]
> [    0.000000]   node   0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
> [    0.000000]   node   0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
> [    0.000000]   node   0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
> [    0.000000]   node   0: [mem 0x0000001fa4030000-0x0000001fa40effff]
> [    0.000000]   node   0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
> [    0.000000]   node   0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
> [    0.000000]   node   0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]
> "
>
>> NORMAL start at 0x1800000000?  And, the free pages will not be scanned
>> there?  Or my understanding were wrong. >
>>> [    0.000000] Movable zone start for each node
>>> [    0.000000] Early memory node ranges
>>> [    0.000000]   node   0: [mem 0x0000000040000000-0x0000000fffffffff]
>>> [    0.000000]   node   0: [mem 0x0000001800000000-0x0000001fa3c7ffff]
>>> [    0.000000]   node   0: [mem 0x0000001fa3c80000-0x0000001fa3ffffff]
>>> [    0.000000]   node   0: [mem 0x0000001fa4000000-0x0000001fa402ffff]
>>> [    0.000000]   node   0: [mem 0x0000001fa4030000-0x0000001fa40effff]
>>> [    0.000000]   node   0: [mem 0x0000001fa40f0000-0x0000001fa73cffff]
>>> [    0.000000]   node   0: [mem 0x0000001fa73d0000-0x0000001fa745ffff]
>>> [    0.000000]   node   0: [mem 0x0000001fa7460000-0x0000001fa746ffff]
>>> [    0.000000]   node   0: [mem 0x0000001fa7470000-0x0000001fa758ffff]
>>> [    0.000000]   node   0: [mem 0x0000001fa7590000-0x0000001fa7ffffff]
>>>
>>> Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
>>> ---
>>>   mm/compaction.c | 30 +++++++++++++++++++++++++++++-
>>>   1 file changed, 29 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/mm/compaction.c b/mm/compaction.c
>>> index 43358efdbdc2..9641e2131901 100644
>>> --- a/mm/compaction.c
>>> +++ b/mm/compaction.c
>>> @@ -249,11 +249,31 @@ static unsigned long skip_offline_sections(unsigned long start_pfn)
>>>     	return 0;
>>>   }
>>> +
>>> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
>>> +{
>>> +	unsigned long start_nr = pfn_to_section_nr(start_pfn);
>>> +
>>> +	if (!start_nr || online_section_nr(start_nr))
>>> +		return 0;
>>> +
>>> +	while (start_nr-- > 0) {
>>> +		if (online_section_nr(start_nr))
>>> +			return section_nr_to_pfn(start_nr) + PAGES_PER_SECTION - 1;
>>> +	}
>>> +
>>> +	return 0;
>>> +}
>>>   #else
>>>   static unsigned long skip_offline_sections(unsigned long start_pfn)
>>>   {
>>>   	return 0;
>>>   }
>>> +
>>> +static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
>>> +{
>>> +	return 0;
>>> +}
>>>   #endif
>>>     /*
>>> @@ -1668,8 +1688,16 @@ static void isolate_freepages(struct compact_control *cc)
>>>     		page = pageblock_pfn_to_page(block_start_pfn,
>>> block_end_pfn,
>>>   									zone);
>>> -		if (!page)
>>> +		if (!page) {
>>> +			unsigned long next_pfn;
>>> +
>>> +			next_pfn = skip_offline_sections_reverse(block_start_pfn);
>>> +			if (next_pfn)
>>> +				block_start_pfn = max(pageblock_start_pfn(next_pfn),
>>> +						      low_pfn);
>>> +
>>>   			continue;
>>> +		}
>>>     		/* Check the block is suitable for migration */
>>>   		if (!suitable_migration_target(cc, page))
  

Patch

diff --git a/mm/compaction.c b/mm/compaction.c
index 43358efdbdc2..9641e2131901 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -249,11 +249,31 @@  static unsigned long skip_offline_sections(unsigned long start_pfn)
 
 	return 0;
 }
+
+static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
+{
+	unsigned long start_nr = pfn_to_section_nr(start_pfn);
+
+	if (!start_nr || online_section_nr(start_nr))
+		return 0;
+
+	while (start_nr-- > 0) {
+		if (online_section_nr(start_nr))
+			return section_nr_to_pfn(start_nr) + PAGES_PER_SECTION - 1;
+	}
+
+	return 0;
+}
 #else
 static unsigned long skip_offline_sections(unsigned long start_pfn)
 {
 	return 0;
 }
+
+static unsigned long skip_offline_sections_reverse(unsigned long start_pfn)
+{
+	return 0;
+}
 #endif
 
 /*
@@ -1668,8 +1688,16 @@  static void isolate_freepages(struct compact_control *cc)
 
 		page = pageblock_pfn_to_page(block_start_pfn, block_end_pfn,
 									zone);
-		if (!page)
+		if (!page) {
+			unsigned long next_pfn;
+
+			next_pfn = skip_offline_sections_reverse(block_start_pfn);
+			if (next_pfn)
+				block_start_pfn = max(pageblock_start_pfn(next_pfn),
+						      low_pfn);
+
 			continue;
+		}
 
 		/* Check the block is suitable for migration */
 		if (!suitable_migration_target(cc, page))