2021-05-27 12:27:28

by Liu Shixin

Subject: [PATCH] mm/page_alloc: fix counting of managed_pages

The commit f63661566fad (mm/page_alloc.c: clear out zone->lowmem_reserve[]
if the zone is empty) clear out zone->lowmem_reserve[] if zone is empty.
But when zone is not empty and sysctl_lowmem_reserve_ratio[i] is set to zero,
zone_managed_pages(zone) is not counted in the managed_pages either. This is
inconsistent with the description of lowmen_reserve, so fix it.

Fixes: f63661566fad ("mm/page_alloc.c: clear out zone->lowmem_reserve[] if the zone is empty")
Reported-by: yangerkun <[email protected]>
Signed-off-by: Liu Shixin <[email protected]>
---
mm/page_alloc.c | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index aaa1655cf682..49a2efce5a84 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -8061,14 +8061,14 @@ static void setup_per_zone_lowmem_reserve(void)
unsigned long managed_pages = 0;

for (j = i + 1; j < MAX_NR_ZONES; j++) {
- if (clear) {
- zone->lowmem_reserve[j] = 0;
- } else {
- struct zone *upper_zone = &pgdat->node_zones[j];
+ struct zone *upper_zone = &pgdat->node_zones[j];
+
+ managed_pages += zone_managed_pages(upper_zone);

- managed_pages += zone_managed_pages(upper_zone);
+ if (clear)
+ zone->lowmem_reserve[j] = 0;
+ else
zone->lowmem_reserve[j] = managed_pages / ratio;
- }
}
}
}
--
2.18.0.huawei.25

2021-05-27 13:17:56

by Liu Shixin

Subject: Re: [PATCH] mm/page_alloc: fix counting of managed_pages

+ Baoquan


On 2021/5/27 20:57, Liu Shixin wrote:
> The commit f63661566fad (mm/page_alloc.c: clear out zone->lowmem_reserve[]
> if the zone is empty) clear out zone->lowmem_reserve[] if zone is empty.
> But when zone is not empty and sysctl_lowmem_reserve_ratio[i] is set to zero,
> zone_managed_pages(zone) is not counted in the managed_pages either. This is
> inconsistent with the description of lowmen_reserve, so fix it.
>
> Fixes: f63661566fad ("mm/page_alloc.c: clear out zone->lowmem_reserve[] if the zone is empty")
> Reported-by: yangerkun <[email protected]>
> Signed-off-by: Liu Shixin <[email protected]>
> ---
> mm/page_alloc.c | 12 ++++++------
> 1 file changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index aaa1655cf682..49a2efce5a84 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -8061,14 +8061,14 @@ static void setup_per_zone_lowmem_reserve(void)
> unsigned long managed_pages = 0;
>
> for (j = i + 1; j < MAX_NR_ZONES; j++) {
> - if (clear) {
> - zone->lowmem_reserve[j] = 0;
> - } else {
> - struct zone *upper_zone = &pgdat->node_zones[j];
> + struct zone *upper_zone = &pgdat->node_zones[j];
> +
> + managed_pages += zone_managed_pages(upper_zone);
>
> - managed_pages += zone_managed_pages(upper_zone);
> + if (clear)
> + zone->lowmem_reserve[j] = 0;
> + else
> zone->lowmem_reserve[j] = managed_pages / ratio;
> - }
> }
> }
> }

2021-05-28 03:19:49

by Baoquan He

Subject: Re: [PATCH] mm/page_alloc: fix counting of managed_pages

On 05/27/21 at 08:57pm, Liu Shixin wrote:
> The commit f63661566fad (mm/page_alloc.c: clear out zone->lowmem_reserve[]
> if the zone is empty) clear out zone->lowmem_reserve[] if zone is empty.
> But when zone is not empty and sysctl_lowmem_reserve_ratio[i] is set to zero,
> zone_managed_pages(zone) is not counted in the managed_pages either. This is
> inconsistent with the description of lowmen_reserve, so fix it.

Right, this is a good catch. Thanks for the fixing.

Is it worth adding words in Documentation/admin-guide/sysctl/vm.rst
to notice this?

Reviewed-by: Baoquan He <[email protected]>

Thanks
Baoquan

>
> Fixes: f63661566fad ("mm/page_alloc.c: clear out zone->lowmem_reserve[] if the zone is empty")
> Reported-by: yangerkun <[email protected]>
> Signed-off-by: Liu Shixin <[email protected]>
> ---
> mm/page_alloc.c | 12 ++++++------
> 1 file changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index aaa1655cf682..49a2efce5a84 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -8061,14 +8061,14 @@ static void setup_per_zone_lowmem_reserve(void)
> unsigned long managed_pages = 0;
>
> for (j = i + 1; j < MAX_NR_ZONES; j++) {
> - if (clear) {
> - zone->lowmem_reserve[j] = 0;
> - } else {
> - struct zone *upper_zone = &pgdat->node_zones[j];
> + struct zone *upper_zone = &pgdat->node_zones[j];
> +
> + managed_pages += zone_managed_pages(upper_zone);
>
> - managed_pages += zone_managed_pages(upper_zone);
> + if (clear)
> + zone->lowmem_reserve[j] = 0;
> + else
> zone->lowmem_reserve[j] = managed_pages / ratio;
> - }
> }
> }
> }
> --
> 2.18.0.huawei.25
>
>

2021-05-28 05:53:52

by Liu Shixin

Subject: Re: [PATCH] mm/page_alloc: fix counting of managed_pages

On 2021/5/28 7:47, Baoquan He wrote:
> On 05/27/21 at 08:57pm, Liu Shixin wrote:
>> The commit f63661566fad (mm/page_alloc.c: clear out zone->lowmem_reserve[]
>> if the zone is empty) clear out zone->lowmem_reserve[] if zone is empty.
>> But when zone is not empty and sysctl_lowmem_reserve_ratio[i] is set to zero,
>> zone_managed_pages(zone) is not counted in the managed_pages either. This is
>> inconsistent with the description of lowmen_reserve, so fix it.
> Right, this is a good catch. Thanks for the fixing.
>
> Is it worth adding words in Documentation/admin-guide/sysctl/vm.rst
> to notice this?
>
> Reviewed-by: Baoquan He <[email protected]>
>
> Thanks
> Baoquan

With this patch, It's consistent with the calculation description of zone[i]'s protection[j].

This is implicit in the description so I don't think it's necessary to notice this.

>> Fixes: f63661566fad ("mm/page_alloc.c: clear out zone->lowmem_reserve[] if the zone is empty")
>> Reported-by: yangerkun <[email protected]>
>> Signed-off-by: Liu Shixin <[email protected]>
>> ---
>> mm/page_alloc.c | 12 ++++++------
>> 1 file changed, 6 insertions(+), 6 deletions(-)
>>
>> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
>> index aaa1655cf682..49a2efce5a84 100644
>> --- a/mm/page_alloc.c
>> +++ b/mm/page_alloc.c
>> @@ -8061,14 +8061,14 @@ static void setup_per_zone_lowmem_reserve(void)
>> unsigned long managed_pages = 0;
>>
>> for (j = i + 1; j < MAX_NR_ZONES; j++) {
>> - if (clear) {
>> - zone->lowmem_reserve[j] = 0;
>> - } else {
>> - struct zone *upper_zone = &pgdat->node_zones[j];
>> + struct zone *upper_zone = &pgdat->node_zones[j];
>> +
>> + managed_pages += zone_managed_pages(upper_zone);
>>
>> - managed_pages += zone_managed_pages(upper_zone);
>> + if (clear)
>> + zone->lowmem_reserve[j] = 0;
>> + else
>> zone->lowmem_reserve[j] = managed_pages / ratio;
>> - }
>> }
>> }
>> }
>> --
>> 2.18.0.huawei.25
>>
>>
> .
>

2021-06-09 14:25:06

by David Hildenbrand

Subject: Re: [PATCH] mm/page_alloc: fix counting of managed_pages

On 27.05.21 14:57, Liu Shixin wrote:
> The commit f63661566fad (mm/page_alloc.c: clear out zone->lowmem_reserve[]
> if the zone is empty) clear out zone->lowmem_reserve[] if zone is empty.
> But when zone is not empty and sysctl_lowmem_reserve_ratio[i] is set to zero,
> zone_managed_pages(zone) is not counted in the managed_pages either. This is
> inconsistent with the description of lowmen_reserve, so fix it.
>
> Fixes: f63661566fad ("mm/page_alloc.c: clear out zone->lowmem_reserve[] if the zone is empty")
> Reported-by: yangerkun <[email protected]>
> Signed-off-by: Liu Shixin <[email protected]>
> ---
> mm/page_alloc.c | 12 ++++++------
> 1 file changed, 6 insertions(+), 6 deletions(-)
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index aaa1655cf682..49a2efce5a84 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -8061,14 +8061,14 @@ static void setup_per_zone_lowmem_reserve(void)
> unsigned long managed_pages = 0;
>
> for (j = i + 1; j < MAX_NR_ZONES; j++) {
> - if (clear) {
> - zone->lowmem_reserve[j] = 0;
> - } else {
> - struct zone *upper_zone = &pgdat->node_zones[j];
> + struct zone *upper_zone = &pgdat->node_zones[j];
> +
> + managed_pages += zone_managed_pages(upper_zone);
>
> - managed_pages += zone_managed_pages(upper_zone);
> + if (clear)
> + zone->lowmem_reserve[j] = 0;
> + else
> zone->lowmem_reserve[j] = managed_pages / ratio;
> - }
> }
> }
> }
>

Acked-by: David Hildenbrand <[email protected]>

--
Thanks,

David / dhildenb