Skip to contentSkip to conversation

mm/arch: Provide pud_pfn() fallback

PATCH In review [email protected] 3 participants 5 messages Mar 23, 2024 – Mar 26, 2024

Subsystems (7)
Memory management x86 MM core RISC-V +3 more subsystems
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm/arch: Provide pud_pfn() fallback

From: Peter Xu <[email protected]>

The comment in the code explains the reasons. We took a different approach comparing to pmd_pfn() by providing a fallback function.

Another option is to provide some lower level config options (compare to HUGETLB_PAGE or THP) to identify which layer an arch can support for such huge mappings. However that can be an overkill.

---

Andrew,

If we care about per-commit build errors (and if it is ever feasible to reorder), we can move this patch to be before the patch "mm/gup: handle huge pud for follow_pud_mask()" in mm-unstable to unbreak build on that commit.

Thanks,

---
 arch/riscv/include/asm/pgtable.h    |  1 +
 arch/s390/include/asm/pgtable.h     |  1 +
 arch/sparc/include/asm/pgtable_64.h |  1 +
 arch/x86/include/asm/pgtable.h      |  1 +
 include/linux/pgtable.h             | 10 ++++++++++
 5 files changed, 14 insertions(+)
arch/riscv/include/asm/pgtable.h +1 -0
index 20242402fc11..0ca28cc8e3fa 100644
@@ -646,6 +646,7 @@ static inline unsigned long pmd_pfn(pmd_t pmd)
 
 #define __pud_to_phys(pud)  (__page_val_to_pfn(pud_val(pud)) << PAGE_SHIFT)
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	return ((__pud_to_phys(pud) & PUD_MASK) >> PAGE_SHIFT);
arch/s390/include/asm/pgtable.h +1 -0
index 1a71cb19c089..6cbbe473f680 100644
@@ -1414,6 +1414,7 @@ static inline unsigned long pud_deref(pud_t pud)
 	return (unsigned long)__va(pud_val(pud) & origin_mask);
 }
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
arch/sparc/include/asm/pgtable_64.h +1 -0
index 4d1bafaba942..26efc9bb644a 100644
@@ -875,6 +875,7 @@ static inline bool pud_leaf(pud_t pud)
 	return pte_val(pte) & _PAGE_PMD_HUGE;
 }
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	pte_t pte = __pte(pud_val(pud));
arch/x86/include/asm/pgtable.h +1 -0
index cefc7a84f7a4..273f7557218c 100644
@@ -234,6 +234,7 @@ static inline unsigned long pmd_pfn(pmd_t pmd)
 	return (pfn & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
 }
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	phys_addr_t pfn = pud_val(pud);
include/linux/pgtable.h +10 -0
index 2a1c044ae467..deae9e50f1a8 100644
@@ -1817,6 +1817,16 @@ typedef unsigned int pgtbl_mod_mask;
 #define pte_leaf_size(x) PAGE_SIZE
 #endif
 
+/*
+ * We always define pmd_pfn for all archs as it's used in lots of generic
+ * code.  Now it happens too for pud_pfn (and can happen for larger
+ * mappings too in the future; we're not there yet).  Instead of defining
+ * it for all archs (like pmd_pfn), provide a fallback.
+ */
+#ifndef pud_pfn
+#define pud_pfn(x) ({ BUILD_BUG(); 0; })
+#endif
+
 /*
  * Some architectures have MMUs that are configurable or selectable at boot
  * time. These lead to variable PTRS_PER_x. For statically allocated arrays it
Signature · 1 line
2.44.0
author
#1 lore raw
Quoted 5 lines · On Sat, Mar 23, 2024 at 11:16:43AM -0400, [email protected] wrote:

From: Peter Xu <[email protected]>

The comment in the code explains the reasons. We took a different approach comparing to pmd_pfn() by providing a fallback function.

Quoted 8 lines

Another option is to provide some lower level config options (compare to HUGETLB_PAGE or THP) to identify which layer an arch can support for such huge mappings. However that can be an overkill.

Also:

Quoted 91 lines

---

Andrew,

If we care about per-commit build errors (and if it is ever feasible to reorder), we can move this patch to be before the patch "mm/gup: handle huge pud for follow_pud_mask()" in mm-unstable to unbreak build on that commit.

Thanks,

---
 arch/riscv/include/asm/pgtable.h    |  1 +
 arch/s390/include/asm/pgtable.h     |  1 +
 arch/sparc/include/asm/pgtable_64.h |  1 +
 arch/x86/include/asm/pgtable.h      |  1 +
 include/linux/pgtable.h             | 10 ++++++++++
 5 files changed, 14 insertions(+)
arch/riscv/include/asm/pgtable.h +1 -0
index 20242402fc11..0ca28cc8e3fa 100644
@@ -646,6 +646,7 @@ static inline unsigned long pmd_pfn(pmd_t pmd)
 
 #define __pud_to_phys(pud)  (__page_val_to_pfn(pud_val(pud)) << PAGE_SHIFT)
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	return ((__pud_to_phys(pud) & PUD_MASK) >> PAGE_SHIFT);
arch/s390/include/asm/pgtable.h +1 -0
index 1a71cb19c089..6cbbe473f680 100644
@@ -1414,6 +1414,7 @@ static inline unsigned long pud_deref(pud_t pud)
 	return (unsigned long)__va(pud_val(pud) & origin_mask);
 }
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	return __pa(pud_deref(pud)) >> PAGE_SHIFT;
arch/sparc/include/asm/pgtable_64.h +1 -0
index 4d1bafaba942..26efc9bb644a 100644
@@ -875,6 +875,7 @@ static inline bool pud_leaf(pud_t pud)
 	return pte_val(pte) & _PAGE_PMD_HUGE;
 }
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	pte_t pte = __pte(pud_val(pud));
arch/x86/include/asm/pgtable.h +1 -0
index cefc7a84f7a4..273f7557218c 100644
@@ -234,6 +234,7 @@ static inline unsigned long pmd_pfn(pmd_t pmd)
 	return (pfn & pmd_pfn_mask(pmd)) >> PAGE_SHIFT;
 }
 
+#define pud_pfn pud_pfn
 static inline unsigned long pud_pfn(pud_t pud)
 {
 	phys_addr_t pfn = pud_val(pud);
include/linux/pgtable.h +10 -0
index 2a1c044ae467..deae9e50f1a8 100644
@@ -1817,6 +1817,16 @@ typedef unsigned int pgtbl_mod_mask;
 #define pte_leaf_size(x) PAGE_SIZE
 #endif
 
+/*
+ * We always define pmd_pfn for all archs as it's used in lots of generic
+ * code.  Now it happens too for pud_pfn (and can happen for larger
+ * mappings too in the future; we're not there yet).  Instead of defining
+ * it for all archs (like pmd_pfn), provide a fallback.
+ */
+#ifndef pud_pfn
+#define pud_pfn(x) ({ BUILD_BUG(); 0; })
+#endif
+
 /*
  * Some architectures have MMUs that are configurable or selectable at boot
  * time. These lead to variable PTRS_PER_x. For statically allocated arrays it
Signature · 1 line
2.44.0
Signature · 1 line
Peter Xu
Reviewed-by on v1 1/1
#2 lore raw
In reply to [email protected]
Quoted 9 lines · On Sat, Mar 23, 2024 at 11:16:43AM -0400, [email protected] wrote:

From: Peter Xu <[email protected]>

The comment in the code explains the reasons. We took a different approach comparing to pmd_pfn() by providing a fallback function.

Another option is to provide some lower level config options (compare to HUGETLB_PAGE or THP) to identify which layer an arch can support for such huge mappings. However that can be an overkill.

Quoted 6 lines

---

Jason

maintainer
#3 lore raw
In reply to [email protected]
Quoted 9 lines · On Sat, 23 Mar 2024 11:16:43 -0400 [email protected] wrote:

From: Peter Xu <[email protected]>

The comment in the code explains the reasons. We took a different approach comparing to pmd_pfn() by providing a fallback function.

Another option is to provide some lower level config options (compare to HUGETLB_PAGE or THP) to identify which layer an arch can support for such huge mappings. However that can be an overkill.

Quoted 6 lines

...

If we care about per-commit build errors (and if it is ever feasible to reorder), we can move this patch to be before the patch "mm/gup: handle huge pud for follow_pud_mask()" in mm-unstable to unbreak build on that commit.

I temporarily disabled that whole series a few days ago. Because of multiple build issues, iirc.

Let's make that permanent. Please redo the whole series against mm-unstable and resend?

author
#4 lore raw
Quoted 18 lines · Andrew Morton
Quoted 9 lines · On Sat, 23 Mar 2024 11:16:43 -0400 [email protected] wrote:

From: Peter Xu <[email protected]>

The comment in the code explains the reasons. We took a different approach comparing to pmd_pfn() by providing a fallback function.

Another option is to provide some lower level config options (compare to HUGETLB_PAGE or THP) to identify which layer an arch can support for such huge mappings. However that can be an overkill.

Quoted 6 lines

...

If we care about per-commit build errors (and if it is ever feasible to reorder), we can move this patch to be before the patch "mm/gup: handle huge pud for follow_pud_mask()" in mm-unstable to unbreak build on that commit.

Quoted 5 lines

I temporarily disabled that whole series a few days ago. Because of multiple build issues, iirc.

Let's make that permanent. Please redo the whole series against mm-unstable and resend?

Yes, that's the plan. Feel free to ignore this as this is not used until that GUP rework series, I'll include it in the whole set to be reposted.

I'm currently doing the build tests; just finished writting the harness for testing the matrix. It'll take a bit time to run through the tests I specified (I tried to cover a few more archs/configs), and I'll repost with all patches included (fixups squashed) when test finished.

[side note: I think I can reproduce the other not-reported issue, on
arm+alldefconfig; that'll get covered too]

Thanks,

Signature · 1 line
Peter Xu