2021-01-21 05:33:43

by Sudarshan Rajagopalan

[permalink] [raw]
Subject: [PATCH 0/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

This patch is the follow-up from the discussions in the thread [1].
Reducing the section size has the merit of reducing wastage of reserved memory
for vmmemmap mappings for sections with large memory holes. Also with smaller
section size gives more grunularity and agility for memory hot(un)plugging.

But there are also constraints in reducing SECTION_SIZE_BIT:

- Should accommodate highest order page for a given config
- Should not break PMD mapping in vmemmap for 4K pages
- Should not consume too many page->flags bits reducing space for other info

This patch uses the suggestions from Anshuman Khandual and David Hildenbrand
in thread [1] to set the least possible section size to 128MB for 4K and 16K
base page size configs for simplicity, and to 512MB for 64K base page size config.

[1] https://lore.kernel.org/lkml/[email protected]/T/#m8ee60ae69db5e9eb06ca7999c43828d49ccb9626


Sudarshan Rajagopalan (1):
arm64/sparsemem: reduce SECTION_SIZE_BITS

arch/arm64/include/asm/sparsemem.h | 23 +++++++++++++++++++++--
1 file changed, 21 insertions(+), 2 deletions(-)

--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project


2021-01-21 05:34:28

by Sudarshan Rajagopalan

[permalink] [raw]
Subject: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

memory_block_size_bytes() determines the memory hotplug granularity i.e the
amount of memory which can be hot added or hot removed from the kernel. The
generic value here being MIN_MEMORY_BLOCK_SIZE (1UL << SECTION_SIZE_BITS)
for memory_block_size_bytes() on platforms like arm64 that does not override.

Current SECTION_SIZE_BITS is 30 i.e 1GB which is large and a reduction here
increases memory hotplug granularity, thus improving its agility. A reduced
section size also reduces memory wastage in vmemmmap mapping for sections
with large memory holes. So we try to set the least section size as possible.

A section size bits selection must follow:
(MAX_ORDER - 1 + PAGE_SHIFT) <= SECTION_SIZE_BITS

CONFIG_FORCE_MAX_ZONEORDER is always defined on arm64 and so just following it
would help achieve the smallest section size.

SECTION_SIZE_BITS = (CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT)

SECTION_SIZE_BITS = 22 (11 - 1 + 12) i.e 4MB for 4K pages
SECTION_SIZE_BITS = 24 (11 - 1 + 14) i.e 16MB for 16K pages without THP
SECTION_SIZE_BITS = 25 (12 - 1 + 14) i.e 32MB for 16K pages with THP
SECTION_SIZE_BITS = 26 (11 - 1 + 16) i.e 64MB for 64K pages without THP
SECTION_SIZE_BITS = 29 (14 - 1 + 16) i.e 512MB for 64K pages with THP

But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
much would over populate /sys/devices/system/memory/ and also consume too many
page->flags bits in the !vmemmap case. Also section size needs to be multiple
of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.

Given these constraints, lets just reduce the section size to 128MB for 4K
and 16K base page size configs, and to 512MB for 64K base page size config.

Signed-off-by: Sudarshan Rajagopalan <[email protected]>
Suggested-by: Anshuman Khandual <[email protected]>
Suggested-by: David Hildenbrand <[email protected]>
Cc: Catalin Marinas <[email protected]>
Cc: Will Deacon <[email protected]>
Cc: Anshuman Khandual <[email protected]>
Cc: David Hildenbrand <[email protected]>
Cc: Mike Rapoport <[email protected]>
Cc: Mark Rutland <[email protected]>
Cc: Logan Gunthorpe <[email protected]>
Cc: Andrew Morton <[email protected]>
Cc: Steven Price <[email protected]>
Cc: Suren Baghdasaryan <[email protected]>
---
arch/arm64/include/asm/sparsemem.h | 23 +++++++++++++++++++++--
1 file changed, 21 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/include/asm/sparsemem.h b/arch/arm64/include/asm/sparsemem.h
index 1f43fcc79738..eb4a75d720ed 100644
--- a/arch/arm64/include/asm/sparsemem.h
+++ b/arch/arm64/include/asm/sparsemem.h
@@ -7,7 +7,26 @@

#ifdef CONFIG_SPARSEMEM
#define MAX_PHYSMEM_BITS CONFIG_ARM64_PA_BITS
-#define SECTION_SIZE_BITS 30
-#endif
+
+/*
+ * Section size must be at least 512MB for 64K base
+ * page size config. Otherwise it will be less than
+ * (MAX_ORDER - 1) and the build process will fail.
+ */
+#ifdef CONFIG_ARM64_64K_PAGES
+#define SECTION_SIZE_BITS 29
+
+#else
+
+/*
+ * Section size must be at least 128MB for 4K base
+ * page size config. Otherwise PMD based huge page
+ * entries could not be created for vmemmap mappings.
+ * 16K follows 4K for simplicity.
+ */
+#define SECTION_SIZE_BITS 27
+#endif /* CONFIG_ARM64_64K_PAGES */
+
+#endif /* CONFIG_SPARSEMEM*/

#endif
--
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project

Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On Wed, 20 Jan 2021, Sudarshan Rajagopalan wrote:

> But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
> much would over populate /sys/devices/system/memory/ and also consume too many
> page->flags bits in the !vmemmap case. Also section size needs to be multiple
> of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.

There is also the issue of requiring more space in the TLB cache with
smaller page sizes. Or does ARM resolve these into smaller TLB entries
anyways (going on my x86 kwon how here)? Anyways if there are only a few
TLB entries then the effect could
be significant.

2021-01-21 13:46:36

by Will Deacon

[permalink] [raw]
Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On Wed, Jan 20, 2021 at 09:29:13PM -0800, Sudarshan Rajagopalan wrote:
> memory_block_size_bytes() determines the memory hotplug granularity i.e the
> amount of memory which can be hot added or hot removed from the kernel. The
> generic value here being MIN_MEMORY_BLOCK_SIZE (1UL << SECTION_SIZE_BITS)
> for memory_block_size_bytes() on platforms like arm64 that does not override.
>
> Current SECTION_SIZE_BITS is 30 i.e 1GB which is large and a reduction here
> increases memory hotplug granularity, thus improving its agility. A reduced
> section size also reduces memory wastage in vmemmmap mapping for sections
> with large memory holes. So we try to set the least section size as possible.
>
> A section size bits selection must follow:
> (MAX_ORDER - 1 + PAGE_SHIFT) <= SECTION_SIZE_BITS
>
> CONFIG_FORCE_MAX_ZONEORDER is always defined on arm64 and so just following it
> would help achieve the smallest section size.
>
> SECTION_SIZE_BITS = (CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT)
>
> SECTION_SIZE_BITS = 22 (11 - 1 + 12) i.e 4MB for 4K pages
> SECTION_SIZE_BITS = 24 (11 - 1 + 14) i.e 16MB for 16K pages without THP
> SECTION_SIZE_BITS = 25 (12 - 1 + 14) i.e 32MB for 16K pages with THP
> SECTION_SIZE_BITS = 26 (11 - 1 + 16) i.e 64MB for 64K pages without THP
> SECTION_SIZE_BITS = 29 (14 - 1 + 16) i.e 512MB for 64K pages with THP
>
> But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
> much would over populate /sys/devices/system/memory/ and also consume too many
> page->flags bits in the !vmemmap case. Also section size needs to be multiple
> of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.
>
> Given these constraints, lets just reduce the section size to 128MB for 4K
> and 16K base page size configs, and to 512MB for 64K base page size config.
>
> Signed-off-by: Sudarshan Rajagopalan <[email protected]>
> Suggested-by: Anshuman Khandual <[email protected]>
> Suggested-by: David Hildenbrand <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: Anshuman Khandual <[email protected]>
> Cc: David Hildenbrand <[email protected]>
> Cc: Mike Rapoport <[email protected]>
> Cc: Mark Rutland <[email protected]>
> Cc: Logan Gunthorpe <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: Steven Price <[email protected]>
> Cc: Suren Baghdasaryan <[email protected]>
> ---
> arch/arm64/include/asm/sparsemem.h | 23 +++++++++++++++++++++--
> 1 file changed, 21 insertions(+), 2 deletions(-)

Anshuman -- are you happy with this now?

Will

2021-01-21 13:58:38

by David Hildenbrand

[permalink] [raw]
Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On 21.01.21 06:29, Sudarshan Rajagopalan wrote:
> memory_block_size_bytes() determines the memory hotplug granularity i.e the
> amount of memory which can be hot added or hot removed from the kernel. The
> generic value here being MIN_MEMORY_BLOCK_SIZE (1UL << SECTION_SIZE_BITS)
> for memory_block_size_bytes() on platforms like arm64 that does not override.
>
> Current SECTION_SIZE_BITS is 30 i.e 1GB which is large and a reduction here
> increases memory hotplug granularity, thus improving its agility. A reduced
> section size also reduces memory wastage in vmemmmap mapping for sections
> with large memory holes. So we try to set the least section size as possible.
>
> A section size bits selection must follow:
> (MAX_ORDER - 1 + PAGE_SHIFT) <= SECTION_SIZE_BITS
>
> CONFIG_FORCE_MAX_ZONEORDER is always defined on arm64 and so just following it
> would help achieve the smallest section size.
>
> SECTION_SIZE_BITS = (CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT)
>
> SECTION_SIZE_BITS = 22 (11 - 1 + 12) i.e 4MB for 4K pages
> SECTION_SIZE_BITS = 24 (11 - 1 + 14) i.e 16MB for 16K pages without THP
> SECTION_SIZE_BITS = 25 (12 - 1 + 14) i.e 32MB for 16K pages with THP
> SECTION_SIZE_BITS = 26 (11 - 1 + 16) i.e 64MB for 64K pages without THP
> SECTION_SIZE_BITS = 29 (14 - 1 + 16) i.e 512MB for 64K pages with THP
>
> But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
> much would over populate /sys/devices/system/memory/ and also consume too many
> page->flags bits in the !vmemmap case. Also section size needs to be multiple
> of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.
>
> Given these constraints, lets just reduce the section size to 128MB for 4K
> and 16K base page size configs, and to 512MB for 64K base page size config.
>
> Signed-off-by: Sudarshan Rajagopalan <[email protected]>
> Suggested-by: Anshuman Khandual <[email protected]>
> Suggested-by: David Hildenbrand <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: Anshuman Khandual <[email protected]>
> Cc: David Hildenbrand <[email protected]>
> Cc: Mike Rapoport <[email protected]>
> Cc: Mark Rutland <[email protected]>
> Cc: Logan Gunthorpe <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: Steven Price <[email protected]>
> Cc: Suren Baghdasaryan <[email protected]>
> ---
> arch/arm64/include/asm/sparsemem.h | 23 +++++++++++++++++++++--
> 1 file changed, 21 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/include/asm/sparsemem.h b/arch/arm64/include/asm/sparsemem.h
> index 1f43fcc79738..eb4a75d720ed 100644
> --- a/arch/arm64/include/asm/sparsemem.h
> +++ b/arch/arm64/include/asm/sparsemem.h
> @@ -7,7 +7,26 @@
>
> #ifdef CONFIG_SPARSEMEM
> #define MAX_PHYSMEM_BITS CONFIG_ARM64_PA_BITS
> -#define SECTION_SIZE_BITS 30
> -#endif
> +
> +/*
> + * Section size must be at least 512MB for 64K base
> + * page size config. Otherwise it will be less than
> + * (MAX_ORDER - 1) and the build process will fail.
> + */
> +#ifdef CONFIG_ARM64_64K_PAGES
> +#define SECTION_SIZE_BITS 29
> +
> +#else
> +
> +/*
> + * Section size must be at least 128MB for 4K base
> + * page size config. Otherwise PMD based huge page
> + * entries could not be created for vmemmap mappings.
> + * 16K follows 4K for simplicity.
> + */
> +#define SECTION_SIZE_BITS 27
> +#endif /* CONFIG_ARM64_64K_PAGES */
> +
> +#endif /* CONFIG_SPARSEMEM*/
>
> #endif
>

I'm happy to see this change.

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

--
Thanks,

David / dhildenb

2021-01-21 14:21:16

by Mike Rapoport

[permalink] [raw]
Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On Wed, Jan 20, 2021 at 09:29:13PM -0800, Sudarshan Rajagopalan wrote:
> memory_block_size_bytes() determines the memory hotplug granularity i.e the
> amount of memory which can be hot added or hot removed from the kernel. The
> generic value here being MIN_MEMORY_BLOCK_SIZE (1UL << SECTION_SIZE_BITS)
> for memory_block_size_bytes() on platforms like arm64 that does not override.
>
> Current SECTION_SIZE_BITS is 30 i.e 1GB which is large and a reduction here
> increases memory hotplug granularity, thus improving its agility. A reduced
> section size also reduces memory wastage in vmemmmap mapping for sections
> with large memory holes. So we try to set the least section size as possible.
>
> A section size bits selection must follow:
> (MAX_ORDER - 1 + PAGE_SHIFT) <= SECTION_SIZE_BITS
>
> CONFIG_FORCE_MAX_ZONEORDER is always defined on arm64 and so just following it
> would help achieve the smallest section size.
>
> SECTION_SIZE_BITS = (CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT)
>
> SECTION_SIZE_BITS = 22 (11 - 1 + 12) i.e 4MB for 4K pages
> SECTION_SIZE_BITS = 24 (11 - 1 + 14) i.e 16MB for 16K pages without THP
> SECTION_SIZE_BITS = 25 (12 - 1 + 14) i.e 32MB for 16K pages with THP
> SECTION_SIZE_BITS = 26 (11 - 1 + 16) i.e 64MB for 64K pages without THP
> SECTION_SIZE_BITS = 29 (14 - 1 + 16) i.e 512MB for 64K pages with THP
>
> But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
> much would over populate /sys/devices/system/memory/ and also consume too many
> page->flags bits in the !vmemmap case. Also section size needs to be multiple
> of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.
>
> Given these constraints, lets just reduce the section size to 128MB for 4K
> and 16K base page size configs, and to 512MB for 64K base page size config.
>
> Signed-off-by: Sudarshan Rajagopalan <[email protected]>
> Suggested-by: Anshuman Khandual <[email protected]>
> Suggested-by: David Hildenbrand <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: Anshuman Khandual <[email protected]>
> Cc: David Hildenbrand <[email protected]>
> Cc: Mike Rapoport <[email protected]>
> Cc: Mark Rutland <[email protected]>
> Cc: Logan Gunthorpe <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: Steven Price <[email protected]>
> Cc: Suren Baghdasaryan <[email protected]>

Acked-by: Mike Rapoport <[email protected]>

BTW, after reduction of the section size maybe arm64 should consider opting
out of freeing unused memory map.

This will make David even more happy as this will allow dropping custom
pfn_valid() ;-)

> ---
> arch/arm64/include/asm/sparsemem.h | 23 +++++++++++++++++++++--
> 1 file changed, 21 insertions(+), 2 deletions(-)
>
> diff --git a/arch/arm64/include/asm/sparsemem.h b/arch/arm64/include/asm/sparsemem.h
> index 1f43fcc79738..eb4a75d720ed 100644
> --- a/arch/arm64/include/asm/sparsemem.h
> +++ b/arch/arm64/include/asm/sparsemem.h
> @@ -7,7 +7,26 @@
>
> #ifdef CONFIG_SPARSEMEM
> #define MAX_PHYSMEM_BITS CONFIG_ARM64_PA_BITS
> -#define SECTION_SIZE_BITS 30
> -#endif
> +
> +/*
> + * Section size must be at least 512MB for 64K base
> + * page size config. Otherwise it will be less than
> + * (MAX_ORDER - 1) and the build process will fail.
> + */
> +#ifdef CONFIG_ARM64_64K_PAGES
> +#define SECTION_SIZE_BITS 29
> +
> +#else
> +
> +/*
> + * Section size must be at least 128MB for 4K base
> + * page size config. Otherwise PMD based huge page
> + * entries could not be created for vmemmap mappings.
> + * 16K follows 4K for simplicity.
> + */
> +#define SECTION_SIZE_BITS 27
> +#endif /* CONFIG_ARM64_64K_PAGES */
> +
> +#endif /* CONFIG_SPARSEMEM*/
>
> #endif
> --
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
> a Linux Foundation Collaborative Project
>

--
Sincerely yours,
Mike.

2021-01-21 15:56:27

by Catalin Marinas

[permalink] [raw]
Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On Wed, Jan 20, 2021 at 09:29:13PM -0800, Sudarshan Rajagopalan wrote:
> memory_block_size_bytes() determines the memory hotplug granularity i.e the
> amount of memory which can be hot added or hot removed from the kernel. The
> generic value here being MIN_MEMORY_BLOCK_SIZE (1UL << SECTION_SIZE_BITS)
> for memory_block_size_bytes() on platforms like arm64 that does not override.
>
> Current SECTION_SIZE_BITS is 30 i.e 1GB which is large and a reduction here
> increases memory hotplug granularity, thus improving its agility. A reduced
> section size also reduces memory wastage in vmemmmap mapping for sections
> with large memory holes. So we try to set the least section size as possible.
>
> A section size bits selection must follow:
> (MAX_ORDER - 1 + PAGE_SHIFT) <= SECTION_SIZE_BITS
>
> CONFIG_FORCE_MAX_ZONEORDER is always defined on arm64 and so just following it
> would help achieve the smallest section size.
>
> SECTION_SIZE_BITS = (CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT)
>
> SECTION_SIZE_BITS = 22 (11 - 1 + 12) i.e 4MB for 4K pages
> SECTION_SIZE_BITS = 24 (11 - 1 + 14) i.e 16MB for 16K pages without THP
> SECTION_SIZE_BITS = 25 (12 - 1 + 14) i.e 32MB for 16K pages with THP
> SECTION_SIZE_BITS = 26 (11 - 1 + 16) i.e 64MB for 64K pages without THP
> SECTION_SIZE_BITS = 29 (14 - 1 + 16) i.e 512MB for 64K pages with THP
>
> But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
> much would over populate /sys/devices/system/memory/ and also consume too many
> page->flags bits in the !vmemmap case. Also section size needs to be multiple
> of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.
>
> Given these constraints, lets just reduce the section size to 128MB for 4K
> and 16K base page size configs, and to 512MB for 64K base page size config.
>
> Signed-off-by: Sudarshan Rajagopalan <[email protected]>
> Suggested-by: Anshuman Khandual <[email protected]>
> Suggested-by: David Hildenbrand <[email protected]>
> Cc: Catalin Marinas <[email protected]>
> Cc: Will Deacon <[email protected]>
> Cc: Anshuman Khandual <[email protected]>
> Cc: David Hildenbrand <[email protected]>
> Cc: Mike Rapoport <[email protected]>
> Cc: Mark Rutland <[email protected]>
> Cc: Logan Gunthorpe <[email protected]>
> Cc: Andrew Morton <[email protected]>
> Cc: Steven Price <[email protected]>
> Cc: Suren Baghdasaryan <[email protected]>

Reviewed-by: Catalin Marinas <[email protected]>

2021-01-21 16:00:38

by Catalin Marinas

[permalink] [raw]
Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On Thu, Jan 21, 2021 at 10:08:17AM +0000, Christoph Lameter wrote:
> On Wed, 20 Jan 2021, Sudarshan Rajagopalan wrote:
>
> > But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
> > much would over populate /sys/devices/system/memory/ and also consume too many
> > page->flags bits in the !vmemmap case. Also section size needs to be multiple
> > of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.
>
> There is also the issue of requiring more space in the TLB cache with
> smaller page sizes. Or does ARM resolve these into smaller TLB entries
> anyways (going on my x86 kwon how here)? Anyways if there are only a few
> TLB entries then the effect could
> be significant.

There is indeed more TLB pressure with smaller page sizes but this patch
doesn't change this.

--
Catalin

2021-01-21 16:08:30

by David Hildenbrand

[permalink] [raw]
Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On 21.01.21 15:16, Mike Rapoport wrote:
> On Wed, Jan 20, 2021 at 09:29:13PM -0800, Sudarshan Rajagopalan wrote:
>> memory_block_size_bytes() determines the memory hotplug granularity i.e the
>> amount of memory which can be hot added or hot removed from the kernel. The
>> generic value here being MIN_MEMORY_BLOCK_SIZE (1UL << SECTION_SIZE_BITS)
>> for memory_block_size_bytes() on platforms like arm64 that does not override.
>>
>> Current SECTION_SIZE_BITS is 30 i.e 1GB which is large and a reduction here
>> increases memory hotplug granularity, thus improving its agility. A reduced
>> section size also reduces memory wastage in vmemmmap mapping for sections
>> with large memory holes. So we try to set the least section size as possible.
>>
>> A section size bits selection must follow:
>> (MAX_ORDER - 1 + PAGE_SHIFT) <= SECTION_SIZE_BITS
>>
>> CONFIG_FORCE_MAX_ZONEORDER is always defined on arm64 and so just following it
>> would help achieve the smallest section size.
>>
>> SECTION_SIZE_BITS = (CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT)
>>
>> SECTION_SIZE_BITS = 22 (11 - 1 + 12) i.e 4MB for 4K pages
>> SECTION_SIZE_BITS = 24 (11 - 1 + 14) i.e 16MB for 16K pages without THP
>> SECTION_SIZE_BITS = 25 (12 - 1 + 14) i.e 32MB for 16K pages with THP
>> SECTION_SIZE_BITS = 26 (11 - 1 + 16) i.e 64MB for 64K pages without THP
>> SECTION_SIZE_BITS = 29 (14 - 1 + 16) i.e 512MB for 64K pages with THP
>>
>> But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
>> much would over populate /sys/devices/system/memory/ and also consume too many
>> page->flags bits in the !vmemmap case. Also section size needs to be multiple
>> of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.
>>
>> Given these constraints, lets just reduce the section size to 128MB for 4K
>> and 16K base page size configs, and to 512MB for 64K base page size config.
>>
>> Signed-off-by: Sudarshan Rajagopalan <[email protected]>
>> Suggested-by: Anshuman Khandual <[email protected]>
>> Suggested-by: David Hildenbrand <[email protected]>
>> Cc: Catalin Marinas <[email protected]>
>> Cc: Will Deacon <[email protected]>
>> Cc: Anshuman Khandual <[email protected]>
>> Cc: David Hildenbrand <[email protected]>
>> Cc: Mike Rapoport <[email protected]>
>> Cc: Mark Rutland <[email protected]>
>> Cc: Logan Gunthorpe <[email protected]>
>> Cc: Andrew Morton <[email protected]>
>> Cc: Steven Price <[email protected]>
>> Cc: Suren Baghdasaryan <[email protected]>
>
> Acked-by: Mike Rapoport <[email protected]>
>
> BTW, after reduction of the section size maybe arm64 should consider opting
> out of freeing unused memory map.
>
> This will make David even more happy as this will allow dropping custom
> pfn_valid() ;-)

Mike knows my wildest dreams ;)

--
Thanks,

David / dhildenb

2021-01-21 18:41:02

by Will Deacon

[permalink] [raw]
Subject: Re: [PATCH 0/1] arm64/sparsemem: reduce SECTION_SIZE_BITS

On Wed, 20 Jan 2021 21:29:12 -0800, Sudarshan Rajagopalan wrote:
> This patch is the follow-up from the discussions in the thread [1].
> Reducing the section size has the merit of reducing wastage of reserved memory
> for vmmemmap mappings for sections with large memory holes. Also with smaller
> section size gives more grunularity and agility for memory hot(un)plugging.
>
> But there are also constraints in reducing SECTION_SIZE_BIT:
>
> [...]

Applied to arm64 (for-next/misc), thanks!

[1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS
https://git.kernel.org/arm64/c/f0b13ee23241

Cheers,
--
Will

https://fixes.arm64.dev
https://next.arm64.dev
https://will.arm64.dev

2021-01-22 03:02:48

by Anshuman Khandual

[permalink] [raw]
Subject: Re: [PATCH 1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS



On 1/21/21 7:06 PM, Will Deacon wrote:
> On Wed, Jan 20, 2021 at 09:29:13PM -0800, Sudarshan Rajagopalan wrote:
>> memory_block_size_bytes() determines the memory hotplug granularity i.e the
>> amount of memory which can be hot added or hot removed from the kernel. The
>> generic value here being MIN_MEMORY_BLOCK_SIZE (1UL << SECTION_SIZE_BITS)
>> for memory_block_size_bytes() on platforms like arm64 that does not override.
>>
>> Current SECTION_SIZE_BITS is 30 i.e 1GB which is large and a reduction here
>> increases memory hotplug granularity, thus improving its agility. A reduced
>> section size also reduces memory wastage in vmemmmap mapping for sections
>> with large memory holes. So we try to set the least section size as possible.
>>
>> A section size bits selection must follow:
>> (MAX_ORDER - 1 + PAGE_SHIFT) <= SECTION_SIZE_BITS
>>
>> CONFIG_FORCE_MAX_ZONEORDER is always defined on arm64 and so just following it
>> would help achieve the smallest section size.
>>
>> SECTION_SIZE_BITS = (CONFIG_FORCE_MAX_ZONEORDER - 1 + PAGE_SHIFT)
>>
>> SECTION_SIZE_BITS = 22 (11 - 1 + 12) i.e 4MB for 4K pages
>> SECTION_SIZE_BITS = 24 (11 - 1 + 14) i.e 16MB for 16K pages without THP
>> SECTION_SIZE_BITS = 25 (12 - 1 + 14) i.e 32MB for 16K pages with THP
>> SECTION_SIZE_BITS = 26 (11 - 1 + 16) i.e 64MB for 64K pages without THP
>> SECTION_SIZE_BITS = 29 (14 - 1 + 16) i.e 512MB for 64K pages with THP
>>
>> But there are other problems in reducing SECTION_SIZE_BIT. Reducing it by too
>> much would over populate /sys/devices/system/memory/ and also consume too many
>> page->flags bits in the !vmemmap case. Also section size needs to be multiple
>> of 128MB to have PMD based vmemmap mapping with CONFIG_ARM64_4K_PAGES.
>>
>> Given these constraints, lets just reduce the section size to 128MB for 4K
>> and 16K base page size configs, and to 512MB for 64K base page size config.
>>
>> Signed-off-by: Sudarshan Rajagopalan <[email protected]>
>> Suggested-by: Anshuman Khandual <[email protected]>
>> Suggested-by: David Hildenbrand <[email protected]>
>> Cc: Catalin Marinas <[email protected]>
>> Cc: Will Deacon <[email protected]>
>> Cc: Anshuman Khandual <[email protected]>
>> Cc: David Hildenbrand <[email protected]>
>> Cc: Mike Rapoport <[email protected]>
>> Cc: Mark Rutland <[email protected]>
>> Cc: Logan Gunthorpe <[email protected]>
>> Cc: Andrew Morton <[email protected]>
>> Cc: Steven Price <[email protected]>
>> Cc: Suren Baghdasaryan <[email protected]>
>> ---
>> arch/arm64/include/asm/sparsemem.h | 23 +++++++++++++++++++++--
>> 1 file changed, 21 insertions(+), 2 deletions(-)
>
> Anshuman -- are you happy with this now?

Yes.

A small nit. There are couple of extra lines in the patch which
can be dropped, probably while merging.

Reviewed-by: Anshuman Khandual <[email protected]>

2021-01-29 00:22:17

by Pasha Tatashin

[permalink] [raw]
Subject: Re: Re: [PATCH 0/1] arm64/sparsemem: reduce SECTION_SIZE_BITS



On 1/21/21 1:26 PM, Will Deacon wrote:
> On Wed, 20 Jan 2021 21:29:12 -0800, Sudarshan Rajagopalan wrote:
>> This patch is the follow-up from the discussions in the thread [1].
>> Reducing the section size has the merit of reducing wastage of reserved memory
>> for vmmemmap mappings for sections with large memory holes. Also with smaller
>> section size gives more grunularity and agility for memory hot(un)plugging.
>>
>> But there are also constraints in reducing SECTION_SIZE_BIT:
>>
>> [...]
>
> Applied to arm64 (for-next/misc), thanks!

We have been using 128M for 4K pages for a while, using patch [1] without issues.

One thing that needs to be also modified is makedumpfile [2]- line 76, must update it SECTIONS_SIZE_BITS based on the PAGESIZE.

[1] https://lore.kernel.org/lkml/[email protected]/
[2] https://sourceforge.net/p/makedumpfile/code/ci/master/tree/arch/arm64.c


>
> [1/1] arm64/sparsemem: reduce SECTION_SIZE_BITS
> https://git.kernel.org/arm64/c/f0b13ee23241
>
> Cheers,
>