Skip to contentSkip to conversation

Generalize vmemmap_populate_hugepages to sparc

PATCH Superseded chengkaitao 3 participants 8 messages Jan 11 – Mar 18

Subsystems (3)

v5 of 6 Superseded Latest is v6 →

lore
Share & tools
lore
#0 lore raw

v5 cover Generalize vmemmap_populate_hugepages to sparc

Change sparc's implementation of vmemmap_populate() using vmemmap_populate_hugepages() to streamline the code. Another benefit is that it allows us to eliminate the external declarations of vmemmap_p?d_populate functions and convert them to static functions.

Since vmemmap_populate_hugepages may fallback to vmemmap_populate- _basepages, which differs from sparc's original implementation. During the v1 discussion with Mike Rapoport, sparc uses base pages in the kernel page tables, so it should be able to use them in vmemmap as well. Consequently, no additional special handling is required.

Remove unnecessary whitespace from sparc.

Changes in v5:
- remove [PATCH v4 3/3] sparc: Remove unnecessary whitespace

Changes in v4:
- Add commit message
- Add verification that pmd actually maps a large page in the vmemmap_check_pmd function

Changes in v3:
- Allow sparc to fallback to vmemmap_populate_basepages
- Convert vmemmap_p?d_populate() to static functions
- Split the v1 patch

Changes in v2:
- Revert the whitespace deletions
- Change vmemmap_false_pmd to vmemmap_pte_fallback_allowed

Link to V4:
https://lore.kernel.org/all/[email protected]/
Link to V3:
https://lore.kernel.org/all/[email protected]/
Link to V2:
https://lore.kernel.org/all/[email protected]/
Link to V1:
https://lore.kernel.org/all/[email protected]/

Chengkaitao (2):
sparc: Use vmemmap_populate_hugepages for vmemmap_populate
mm: Convert vmemmap_p?d_populate() to static functions

 arch/sparc/mm/init_64.c | 47 ++++++++++++++---------------------------
 include/linux/mm.h      |  7 ------
 mm/sparse-vmemmap.c     | 10 ++++-----
 3 files changed, 21 insertions(+), 43 deletions(-)
Signature · 1 line
2.50.1 (Apple Git-155)
#1 lore raw

PATCH v5 1/2 sparc: Use vmemmap_populate_hugepages for vmemmap_populate

From: Chengkaitao <[email protected]>

1. In the SPARC architecture, reimplemented vmemmap_populate using vmemmap_populate_hugepages.
2. Allow the SPARC arch to fallback to vmemmap_populate_basepages(), when vmemmap_alloc_block returns NULL.

---
 arch/sparc/mm/init_64.c | 47 ++++++++++++++---------------------------
 1 file changed, 16 insertions(+), 31 deletions(-)
arch/sparc/mm/init_64.c +16 -31
index df9f7c444c39..858eaa6615ea 100644
@@ -2581,8 +2581,8 @@ unsigned long _PAGE_CACHE __read_mostly;
 EXPORT_SYMBOL(_PAGE_CACHE);
 
 #ifdef CONFIG_SPARSEMEM_VMEMMAP
-int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
-			       int node, struct vmem_altmap *altmap)
+void __meminit vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
+			       unsigned long addr, unsigned long next)
 {
 	unsigned long pte_base;
 
@@ -2595,39 +2595,24 @@ int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
 
 	pte_base |= _PAGE_PMD_HUGE;
 
-	vstart = vstart & PMD_MASK;
-	vend = ALIGN(vend, PMD_SIZE);
-	for (; vstart < vend; vstart += PMD_SIZE) {
-		pgd_t *pgd = vmemmap_pgd_populate(vstart, node);
-		unsigned long pte;
-		p4d_t *p4d;
-		pud_t *pud;
-		pmd_t *pmd;
-
-		if (!pgd)
-			return -ENOMEM;
-
-		p4d = vmemmap_p4d_populate(pgd, vstart, node);
-		if (!p4d)
-			return -ENOMEM;
-
-		pud = vmemmap_pud_populate(p4d, vstart, node);
-		if (!pud)
-			return -ENOMEM;
+	pmd_val(*pmd) = pte_base | __pa(p);
+}
 
-		pmd = pmd_offset(pud, vstart);
-		pte = pmd_val(*pmd);
-		if (!(pte & _PAGE_VALID)) {
-			void *block = vmemmap_alloc_block(PMD_SIZE, node);
+int __meminit vmemmap_check_pmd(pmd_t *pmdp, int node,
+				unsigned long addr, unsigned long next)
+{
+	int large = pmd_leaf(*pmdp);
 
-			if (!block)
-				return -ENOMEM;
+	if (large)
+		vmemmap_verify((pte_t *)pmdp, node, addr, next);
 
-			pmd_val(*pmd) = pte_base | __pa(block);
-		}
-	}
+	return large;
+}
 
-	return 0;
+int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
+			       int node, struct vmem_altmap *altmap)
+{
+	return vmemmap_populate_hugepages(vstart, vend, node, altmap);
 }
 #endif /* CONFIG_SPARSEMEM_VMEMMAP */
 
Signature · 1 line
2.50.1 (Apple Git-155)
maintainer
#2 lore raw

Re: [PATCH v5 1/2] sparc: Use vmemmap_populate_hugepages for vmemmap_populate

Quoted 6 lines · chengkaitao

From: Chengkaitao <[email protected]>

1. In the SPARC architecture, reimplemented vmemmap_populate using vmemmap_populate_hugepages.
2. Allow the SPARC arch to fallback to vmemmap_populate_basepages(), when vmemmap_alloc_block returns NULL.

This patch seems to potentially make more functional changes than what the descriptions gives impression of.

Given the amount of changes this seems to introduce, more on that below, I'd like to see more description on the changes and why they can be done than this.

Nit: use active language, "reimplement", not "reimplemented".

Quoted 22 lines
---
 arch/sparc/mm/init_64.c | 47 ++++++++++++++---------------------------
 1 file changed, 16 insertions(+), 31 deletions(-)
arch/sparc/mm/init_64.c +2 -2
index df9f7c444c39..858eaa6615ea 100644
@@ -2581,8 +2581,8 @@ unsigned long _PAGE_CACHE __read_mostly;
 EXPORT_SYMBOL(_PAGE_CACHE);
 
 #ifdef CONFIG_SPARSEMEM_VMEMMAP
-int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
-			       int node, struct vmem_altmap *altmap)
+void __meminit vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
+			       unsigned long addr, unsigned long next)
 {
 	unsigned long pte_base;
 

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2599

Quoted 5 lines
arch/sparc/mm/init_64.c +0 -2
@@ -2595,39 +2595,24 @@ int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
 
 	pte_base |= _PAGE_PMD_HUGE;
 
-	vstart = vstart & PMD_MASK;
-	vend = ALIGN(vend, PMD_SIZE);

It seems that this patch removes alignment of both start and end. Is this a functional change in practice or are these always aligned for some other reason?

Quoted 13 lines
arch/sparc/mm/init_64.c +0 -13
-	for (; vstart < vend; vstart += PMD_SIZE) {
-		pgd_t *pgd = vmemmap_pgd_populate(vstart, node);
-		unsigned long pte;
-		p4d_t *p4d;
-		pud_t *pud;
-		pmd_t *pmd;
-
-		if (!pgd)
-			return -ENOMEM;
-
-		p4d = vmemmap_p4d_populate(pgd, vstart, node);
-		if (!p4d)
-			return -ENOMEM;

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2620

Quoted 10 lines
arch/sparc/mm/init_64.c +2 -7
-
-		pud = vmemmap_pud_populate(p4d, vstart, node);
-		if (!pud)
-			return -ENOMEM;
+	pmd_val(*pmd) = pte_base | __pa(p);
+}
 
-		pmd = pmd_offset(pud, vstart);
-		pte = pmd_val(*pmd);
-		if (!(pte & _PAGE_VALID)) {

It is not the same thing, but is this equivalent to if (pmd_none(pmdp_get(pmd))) at this point?

Quoted 15 lines
arch/sparc/mm/init_64.c +7 -6
-			void *block = vmemmap_alloc_block(PMD_SIZE, node);
+int __meminit vmemmap_check_pmd(pmd_t *pmdp, int node,
+				unsigned long addr, unsigned long next)
+{
+	int large = pmd_leaf(*pmdp);
 
-			if (!block)
-				return -ENOMEM;
+	if (large)
+		vmemmap_verify((pte_t *)pmdp, node, addr, next);
 
-			pmd_val(*pmd) = pte_base | __pa(block);
-		}
-	}
+	return large;

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2618

Quoted 10 lines
arch/sparc/mm/init_64.c +5 -1
+}
 
-	return 0;
+int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
+			       int node, struct vmem_altmap *altmap)
+{
+	return vmemmap_populate_hugepages(vstart, vend, node, altmap);
 }
 #endif /* CONFIG_SPARSEMEM_VMEMMAP */
 

This change introduces using vmemmap_alloc_block_buf() instead of vmemmap_alloc_block() seems to introduce two new behaviours that was not in use for sparc64 before:

1) Using altmap_alloc_block_buf() for a non-null altmap, that was not used before. Also the fallback to vmemmap_populate_basepages() passes on altmap.

2) Trying sparse_buffer_alloc() before vmemmap_alloc_block(), which was not done before.

Neither the commit message nor the cover letter touches upon this. Could you elaborate here?

Given all the (at least seeming) functional changes could you share how you tested this change?

Cheers,
Andreas

maintainer
#3 lore raw

Re: [PATCH v5 1/2] sparc: Use vmemmap_populate_hugepages for vmemmap_populate

Quoted 8 lines · On Mon, 26 Jan 2026 15:50:34 +0100 Andreas Larsson <[email protected]> wrote:
Quoted 6 lines · chengkaitao

From: Chengkaitao <[email protected]>

1. In the SPARC architecture, reimplemented vmemmap_populate using vmemmap_populate_hugepages.
2. Allow the SPARC arch to fallback to vmemmap_populate_basepages(), when vmemmap_alloc_block returns NULL.

Quoted 8 lines

This patch seems to potentially make more functional changes than what the descriptions gives impression of.

Given the amount of changes this seems to introduce, more on that below, I'd like to see more description on the changes and why they can be done than this.

Nit: use active language, "reimplement", not "reimplemented".

Thanks, I'll drop the v5 version of this patchset.

#4 lore raw

Re: [PATCH v5 1/2] sparc: Use vmemmap_populate_hugepages for vmemmap_populate

In reply to Andreas Larsson
Quoted 42 lines · Andreas Larsson
Quoted 6 lines · chengkaitao

From: Chengkaitao <[email protected]>

1. In the SPARC architecture, reimplemented vmemmap_populate using vmemmap_populate_hugepages.
2. Allow the SPARC arch to fallback to vmemmap_populate_basepages(), when vmemmap_alloc_block returns NULL.

This patch seems to potentially make more functional changes than what the descriptions gives impression of.

Given the amount of changes this seems to introduce, more on that below, I'd like to see more description on the changes and why they can be done than this.

Nit: use active language, "reimplement", not "reimplemented".

Quoted 23 lines
---
 arch/sparc/mm/init_64.c | 47 ++++++++++++++---------------------------
 1 file changed, 16 insertions(+), 31 deletions(-)
arch/sparc/mm/init_64.c +2 -2
index df9f7c444c39..858eaa6615ea 100644
@@ -2581,8 +2581,8 @@ unsigned long _PAGE_CACHE __read_mostly;
 EXPORT_SYMBOL(_PAGE_CACHE);
 
 #ifdef CONFIG_SPARSEMEM_VMEMMAP
-int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
-                            int node, struct vmem_altmap *altmap)
+void __meminit vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
+                            unsigned long addr, unsigned long next)
 {
      unsigned long pte_base;
 
@@ -2595,39 +2595,24 @@ int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
 
Quoted 9 lines

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2599

Quoted 4 lines
arch/sparc/mm/init_64.c +0 -2
      pte_base |= _PAGE_PMD_HUGE;
 
-     vstart = vstart & PMD_MASK;
-     vend = ALIGN(vend, PMD_SIZE);

It seems that this patch removes alignment of both start and end. Is this a functional change in practice or are these always aligned for some other reason?

Whether vstart and vend are aligned with PMD_SIZE doesn't seem to affect the behavior pattern or output of vmemmap_populate_hugepages. The vmemmap_populate_hugepages function performs necessary alignment processing internally, such as pmd_addr_end and pmd/pte_index?

Quoted 20 lines
Quoted 20 lines
arch/sparc/mm/init_64.c +2 -17
-     for (; vstart < vend; vstart += PMD_SIZE) {
-             pgd_t *pgd = vmemmap_pgd_populate(vstart, node);
-             unsigned long pte;
-             p4d_t *p4d;
-             pud_t *pud;
-             pmd_t *pmd;
-
-             if (!pgd)
-                     return -ENOMEM;
-
-             p4d = vmemmap_p4d_populate(pgd, vstart, node);
-             if (!p4d)
-                     return -ENOMEM;
-
-             pud = vmemmap_pud_populate(p4d, vstart, node);
-             if (!pud)
-                     return -ENOMEM;
+     pmd_val(*pmd) = pte_base | __pa(p);
+}
 
Quoted 7 lines

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2620

Quoted 3 lines
arch/sparc/mm/init_64.c +0 -3
-             pmd = pmd_offset(pud, vstart);
-             pte = pmd_val(*pmd);
-             if (!(pte & _PAGE_VALID)) {

It is not the same thing, but is this equivalent to if (pmd_none(pmdp_get(pmd))) at this point?

For PMD entries, there shouldn't be cases where pmd_none and _PAGE_VALID exhibit inconsistent behavior. I've observed that pmd_none is widely used in the SPARC architecture.

Quoted 31 lines
Quoted 15 lines
arch/sparc/mm/init_64.c +7 -6
-                     void *block = vmemmap_alloc_block(PMD_SIZE, node);
+int __meminit vmemmap_check_pmd(pmd_t *pmdp, int node,
+                             unsigned long addr, unsigned long next)
+{
+     int large = pmd_leaf(*pmdp);
 
-                     if (!block)
-                             return -ENOMEM;
+     if (large)
+             vmemmap_verify((pte_t *)pmdp, node, addr, next);
 
-                     pmd_val(*pmd) = pte_base | __pa(block);
-             }
-     }
+     return large;

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2618

Quoted 10 lines
arch/sparc/mm/init_64.c +5 -1
+}
 
-     return 0;
+int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
+                            int node, struct vmem_altmap *altmap)
+{
+     return vmemmap_populate_hugepages(vstart, vend, node, altmap);
 }
 #endif /* CONFIG_SPARSEMEM_VMEMMAP */
 

This change introduces using vmemmap_alloc_block_buf() instead of vmemmap_alloc_block() seems to introduce two new behaviours that was not in use for sparc64 before:

Quoted 3 lines

1) Using altmap_alloc_block_buf() for a non-null altmap, that was not used before. Also the fallback to vmemmap_populate_basepages() passes on altmap.

If altmap validation isn't required, I can retain the original code logic by setting altmap to NULL.

Quoted 2 lines

2) Trying sparse_buffer_alloc() before vmemmap_alloc_block(), which was not done before.

In SPARC, sparse_init() is called to initialize the sparsemap_buf. If the SPARC architecture doesn't support using sparse_buffer_alloc, we can remove the sparse_init() call path.

Quoted 5 lines

Neither the commit message nor the cover letter touches upon this. Could you elaborate here?

Given all the (at least seeming) functional changes could you share how you tested this change?

My original intention was to help architectures adopt more generic kernel APIs to reduce maintenance costs. However, due to my lack of physical SPARC devices, I couldn't perform comprehensive testing, I've only verified compilation correctness based on code analysis. I sincerely apologize for this limitation. If you have access to physical SPARC hardware, could you kindly help with testing?

Signature · 2 lines
Cheers,
Chengkaitao
maintainer
#5 lore raw

Re: [PATCH v5 1/2] sparc: Use vmemmap_populate_hugepages for vmemmap_populate

Quoted 126 lines · Chengkaitao
Quoted 42 lines · Andreas Larsson
Quoted 6 lines · chengkaitao

From: Chengkaitao <[email protected]>

1. In the SPARC architecture, reimplemented vmemmap_populate using vmemmap_populate_hugepages.
2. Allow the SPARC arch to fallback to vmemmap_populate_basepages(), when vmemmap_alloc_block returns NULL.

This patch seems to potentially make more functional changes than what the descriptions gives impression of.

Given the amount of changes this seems to introduce, more on that below, I'd like to see more description on the changes and why they can be done than this.

Nit: use active language, "reimplement", not "reimplemented".

Quoted 23 lines
---
 arch/sparc/mm/init_64.c | 47 ++++++++++++++---------------------------
 1 file changed, 16 insertions(+), 31 deletions(-)
arch/sparc/mm/init_64.c +2 -2
index df9f7c444c39..858eaa6615ea 100644
@@ -2581,8 +2581,8 @@ unsigned long _PAGE_CACHE __read_mostly;
 EXPORT_SYMBOL(_PAGE_CACHE);
 
 #ifdef CONFIG_SPARSEMEM_VMEMMAP
-int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
-                            int node, struct vmem_altmap *altmap)
+void __meminit vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
+                            unsigned long addr, unsigned long next)
 {
      unsigned long pte_base;
 
@@ -2595,39 +2595,24 @@ int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
 
Quoted 9 lines

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2599

Quoted 4 lines
arch/sparc/mm/init_64.c +0 -2
      pte_base |= _PAGE_PMD_HUGE;
 
-     vstart = vstart & PMD_MASK;
-     vend = ALIGN(vend, PMD_SIZE);

It seems that this patch removes alignment of both start and end. Is this a functional change in practice or are these always aligned for some other reason?

Whether vstart and vend are aligned with PMD_SIZE doesn't seem to affect the behavior pattern or output of vmemmap_populate_hugepages. The vmemmap_populate_hugepages function performs necessary alignment processing internally, such as pmd_addr_end and pmd/pte_index?

Quoted 20 lines
Quoted 20 lines
arch/sparc/mm/init_64.c +2 -17
-     for (; vstart < vend; vstart += PMD_SIZE) {
-             pgd_t *pgd = vmemmap_pgd_populate(vstart, node);
-             unsigned long pte;
-             p4d_t *p4d;
-             pud_t *pud;
-             pmd_t *pmd;
-
-             if (!pgd)
-                     return -ENOMEM;
-
-             p4d = vmemmap_p4d_populate(pgd, vstart, node);
-             if (!p4d)
-                     return -ENOMEM;
-
-             pud = vmemmap_pud_populate(p4d, vstart, node);
-             if (!pud)
-                     return -ENOMEM;
+     pmd_val(*pmd) = pte_base | __pa(p);
+}
 
Quoted 7 lines

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2620

Quoted 3 lines
arch/sparc/mm/init_64.c +0 -3
-             pmd = pmd_offset(pud, vstart);
-             pte = pmd_val(*pmd);
-             if (!(pte & _PAGE_VALID)) {

It is not the same thing, but is this equivalent to if (pmd_none(pmdp_get(pmd))) at this point?

For PMD entries, there shouldn't be cases where pmd_none and _PAGE_VALID exhibit inconsistent behavior. I've observed that pmd_none is widely used in the SPARC architecture.

Quoted 31 lines
Quoted 15 lines
arch/sparc/mm/init_64.c +7 -6
-                     void *block = vmemmap_alloc_block(PMD_SIZE, node);
+int __meminit vmemmap_check_pmd(pmd_t *pmdp, int node,
+                             unsigned long addr, unsigned long next)
+{
+     int large = pmd_leaf(*pmdp);
 
-                     if (!block)
-                             return -ENOMEM;
+     if (large)
+             vmemmap_verify((pte_t *)pmdp, node, addr, next);
 
-                     pmd_val(*pmd) = pte_base | __pa(block);
-             }
-     }
+     return large;

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2618

Quoted 10 lines
arch/sparc/mm/init_64.c +5 -1
+}
 
-     return 0;
+int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
+                            int node, struct vmem_altmap *altmap)
+{
+     return vmemmap_populate_hugepages(vstart, vend, node, altmap);
 }
 #endif /* CONFIG_SPARSEMEM_VMEMMAP */
 

This change introduces using vmemmap_alloc_block_buf() instead of vmemmap_alloc_block() seems to introduce two new behaviours that was not in use for sparc64 before:

Quoted 3 lines

1) Using altmap_alloc_block_buf() for a non-null altmap, that was not used before. Also the fallback to vmemmap_populate_basepages() passes on altmap.

If altmap validation isn't required, I can retain the original code logic by setting altmap to NULL.

Quoted 6 lines
Quoted 2 lines

2) Trying sparse_buffer_alloc() before vmemmap_alloc_block(), which was not done before.

In SPARC, sparse_init() is called to initialize the sparsemap_buf. If the SPARC architecture doesn't support using sparse_buffer_alloc, we can remove the sparse_init() call path.

Thank you for the details.

Quoted 6 lines
Quoted 5 lines

Neither the commit message nor the cover letter touches upon this. Could you elaborate here?

Given all the (at least seeming) functional changes could you share how you tested this change?

Quoted 6 lines

My original intention was to help architectures adopt more generic kernel APIs to reduce maintenance costs. However, due to my lack of physical SPARC devices, I couldn't perform comprehensive testing, I've only verified compilation correctness based on code analysis. I sincerely apologize for this limitation. If you have access to physical SPARC hardware, could you kindly help with testing?

Yes, I have tested v6 on SPARC hardware.

Cheers,
Andreas

#6 lore raw

Re: [PATCH v5 1/2] sparc: Use vmemmap_populate_hugepages for vmemmap_populate

In reply to Andreas Larsson
Quoted 9 lines · Andreas Larsson
Quoted 6 lines · chengkaitao

From: Chengkaitao <[email protected]>

1. In the SPARC architecture, reimplemented vmemmap_populate using vmemmap_populate_hugepages.
2. Allow the SPARC arch to fallback to vmemmap_populate_basepages(), when vmemmap_alloc_block returns NULL.

Quoted 6 lines

This patch seems to potentially make more functional changes than what the descriptions gives impression of.

Given the amount of changes this seems to introduce, more on that below, I'd like to see more description on the changes and why they can be done than this.

Hi Andreas Larsson,
Regarding my previous responses to the questions raised in this email thread, I have not yet received your reply. I would like to provide additional clarifications on these points here:

Quoted 1 line

Nit: use active language, "reimplement", not "reimplemented".

This issue has been fixed in version v6.
https://lore.kernel.org/all/[email protected]/

Quoted 23 lines
Quoted 23 lines
---
 arch/sparc/mm/init_64.c | 47 ++++++++++++++---------------------------
 1 file changed, 16 insertions(+), 31 deletions(-)
arch/sparc/mm/init_64.c +2 -2
index df9f7c444c39..858eaa6615ea 100644
@@ -2581,8 +2581,8 @@ unsigned long _PAGE_CACHE __read_mostly;
 EXPORT_SYMBOL(_PAGE_CACHE);
 
 #ifdef CONFIG_SPARSEMEM_VMEMMAP
-int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
-                            int node, struct vmem_altmap *altmap)
+void __meminit vmemmap_set_pmd(pmd_t *pmd, void *p, int node,
+                            unsigned long addr, unsigned long next)
 {
      unsigned long pte_base;
 
@@ -2595,39 +2595,24 @@ int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
 
Quoted 8 lines

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2599

Quoted 4 lines
arch/sparc/mm/init_64.c +0 -2
      pte_base |= _PAGE_PMD_HUGE;
 
-     vstart = vstart & PMD_MASK;
-     vend = ALIGN(vend, PMD_SIZE);

It seems that this patch removes alignment of both start and end. Is this a functional change in practice or are these always aligned for some other reason?

In the implementation of vmemmap_populate_hugepages, the start value remains unaligned when entering the for-loop for the first time. However, there is no need to apply the "start & PMD_MASK" operation in the vmemmap_*_populate series of functions. In iterations after the first one, pmd_addr_end() will align the start value; since start is already aligned, there is no longer a need to align end.

In fact, in the original code, the presence of
"vstart = vstart & PMD_MASK" made "vend = ALIGN(vend, PMD_SIZE)" redundant as well.

Quoted 20 lines
Quoted 20 lines
arch/sparc/mm/init_64.c +2 -17
-     for (; vstart < vend; vstart += PMD_SIZE) {
-             pgd_t *pgd = vmemmap_pgd_populate(vstart, node);
-             unsigned long pte;
-             p4d_t *p4d;
-             pud_t *pud;
-             pmd_t *pmd;
-
-             if (!pgd)
-                     return -ENOMEM;
-
-             p4d = vmemmap_p4d_populate(pgd, vstart, node);
-             if (!p4d)
-                     return -ENOMEM;
-
-             pud = vmemmap_pud_populate(p4d, vstart, node);
-             if (!pud)
-                     return -ENOMEM;
+     pmd_val(*pmd) = pte_base | __pa(p);
+}
 
Quoted 6 lines

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2620

Quoted 3 lines
arch/sparc/mm/init_64.c +0 -3
-             pmd = pmd_offset(pud, vstart);
-             pte = pmd_val(*pmd);
-             if (!(pte & _PAGE_VALID)) {

It is not the same thing, but is this equivalent to if (pmd_none(pmdp_get(pmd))) at this point?

Since the vmemmap_*_populate functions all use vmemmap_alloc_block_zero when allocating page tables, (pmd_none(pmdp_get(pmd))) can be used as a replacement at this point.

Quoted 31 lines
Quoted 15 lines
arch/sparc/mm/init_64.c +7 -6
-                     void *block = vmemmap_alloc_block(PMD_SIZE, node);
+int __meminit vmemmap_check_pmd(pmd_t *pmdp, int node,
+                             unsigned long addr, unsigned long next)
+{
+     int large = pmd_leaf(*pmdp);
 
-                     if (!block)
-                             return -ENOMEM;
+     if (large)
+             vmemmap_verify((pte_t *)pmdp, node, addr, next);
 
-                     pmd_val(*pmd) = pte_base | __pa(block);
-             }
-     }
+     return large;

arch/sparc/mm/init_64.c int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend, · line 2618

Quoted 10 lines
arch/sparc/mm/init_64.c +5 -1
+}
 
-     return 0;
+int __meminit vmemmap_populate(unsigned long vstart, unsigned long vend,
+                            int node, struct vmem_altmap *altmap)
+{
+     return vmemmap_populate_hugepages(vstart, vend, node, altmap);
 }
 #endif /* CONFIG_SPARSEMEM_VMEMMAP */
 

This change introduces using vmemmap_alloc_block_buf() instead of vmemmap_alloc_block() seems to introduce two new behaviours that was not in use for sparc64 before:

Quoted 3 lines

1) Using altmap_alloc_block_buf() for a non-null altmap, that was not used before. Also the fallback to vmemmap_populate_basepages() passes on altmap.

This issue has been fixed in version v6.
https://lore.kernel.org/all/[email protected]/

Quoted 2 lines

2) Trying sparse_buffer_alloc() before vmemmap_alloc_block(), which was not done before.

The sparse_buffer_alloc function here is usable.
Both sparse_buffer_alloc() and vmemmap_alloc_block() essentially call memmap_alloc(), the only difference is that sparse_buffer_alloc performs the allocation in advance.

Quoted 9 lines

Neither the commit message nor the cover letter touches upon this. Could you elaborate here?

Given all the (at least seeming) functional changes could you share how you tested this change?

Cheers,
Andreas

Is there still a possibility for this patch to be merged into the mainline? Should I continue modifying and iterating on it? Could you help with testing it?

Regardless of your answer, I look forward to your reply.

Signature · 2 lines
Yours,
Chengkaitao
#7 lore raw

PATCH v5 2/2 mm: Convert vmemmap_p?d_populate() to static functions

In reply to chengkaitao

From: Chengkaitao <[email protected]>

Since the vmemmap_p?d_populate functions are unused outside the mm subsystem, we can remove their external declarations and convert them to static functions.

---
 include/linux/mm.h  |  7 -------
 mm/sparse-vmemmap.c | 10 +++++-----
 2 files changed, 5 insertions(+), 12 deletions(-)
include/linux/mm.h +0 -7
index 6f959d8ca4b4..322fe72b6ebd 100644
@@ -4234,13 +4234,6 @@ unsigned long section_map_size(void);
 struct page * __populate_section_memmap(unsigned long pfn,
 		unsigned long nr_pages, int nid, struct vmem_altmap *altmap,
 		struct dev_pagemap *pgmap);
-pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
-p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
-pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
-pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
-pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
-			    struct vmem_altmap *altmap, unsigned long ptpfn,
-			    unsigned long flags);
 void *vmemmap_alloc_block(unsigned long size, int node);
 struct vmem_altmap;
 void *vmemmap_alloc_block_buf(unsigned long size, int node,
mm/sparse-vmemmap.c +5 -5
index 37522d6cb398..67f9a08c2376 100644
@@ -151,7 +151,7 @@ void __meminit vmemmap_verify(pte_t *pte, int node,
 			start, end - 1);
 }
 
-pte_t * __meminit vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
+static pte_t * __meminit vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node,
 				       struct vmem_altmap *altmap,
 				       unsigned long ptpfn, unsigned long flags)
 {
@@ -195,7 +195,7 @@ static void * __meminit vmemmap_alloc_block_zero(unsigned long size, int node)
 	return p;
 }
 
-pmd_t * __meminit vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node)
+static pmd_t * __meminit vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node)
 {
 	pmd_t *pmd = pmd_offset(pud, addr);
 	if (pmd_none(*pmd)) {
@@ -208,7 +208,7 @@ pmd_t * __meminit vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node)
 	return pmd;
 }
 
-pud_t * __meminit vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node)
+static pud_t * __meminit vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node)
 {
 	pud_t *pud = pud_offset(p4d, addr);
 	if (pud_none(*pud)) {
@@ -221,7 +221,7 @@ pud_t * __meminit vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node)
 	return pud;
 }
 
-p4d_t * __meminit vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node)
+static p4d_t * __meminit vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node)
 {
 	p4d_t *p4d = p4d_offset(pgd, addr);
 	if (p4d_none(*p4d)) {
@@ -234,7 +234,7 @@ p4d_t * __meminit vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node)
 	return p4d;
 }
 
-pgd_t * __meminit vmemmap_pgd_populate(unsigned long addr, int node)
+static pgd_t * __meminit vmemmap_pgd_populate(unsigned long addr, int node)
 {
 	pgd_t *pgd = pgd_offset_k(addr);
 	if (pgd_none(*pgd)) {
Signature · 1 line
2.50.1 (Apple Git-155)