kvm: Use huge pages for DAX-backed files

KVM has a use case for determining the size of a dax mapping. The KVM code has easy access to the address and the mm; hence the change in parameters.

Signed-off-by Barret Rhoden <[email protected]>

---
 include/linux/mm.h  |  3 +++
 mm/memory-failure.c | 38 +++-----------------------------------
 mm/util.c           | 34 ++++++++++++++++++++++++++++++++++
 3 files changed, 40 insertions(+), 35 deletions(-)

include/linux/mm.h +3 -0

index 5411de93a363..51215d695753 100644
@@ -935,6 +935,9 @@ static inline bool is_pci_p2pdma_page(const struct page *page)
 }
 #endif /* CONFIG_DEV_PAGEMAP_OPS */
 
+unsigned long dev_pagemap_mapping_shift(unsigned long address,
+					struct mm_struct *mm);
+
 static inline void get_page(struct page *page)
 {
 	page = compound_head(page);

mm/memory-failure.c +3 -35

index 0cd3de3550f0..c3f2c6a8607e 100644
@@ -265,40 +265,6 @@ void shake_page(struct page *p, int access)
 }
 EXPORT_SYMBOL_GPL(shake_page);
 
-static unsigned long dev_pagemap_mapping_shift(struct page *page,
-		struct vm_area_struct *vma)
-{
-	unsigned long address = vma_address(page, vma);
-	pgd_t *pgd;
-	p4d_t *p4d;
-	pud_t *pud;
-	pmd_t *pmd;
-	pte_t *pte;
-
-	pgd = pgd_offset(vma->vm_mm, address);
-	if (!pgd_present(*pgd))
-		return 0;
-	p4d = p4d_offset(pgd, address);
-	if (!p4d_present(*p4d))
-		return 0;
-	pud = pud_offset(p4d, address);
-	if (!pud_present(*pud))
-		return 0;
-	if (pud_devmap(*pud))
-		return PUD_SHIFT;
-	pmd = pmd_offset(pud, address);
-	if (!pmd_present(*pmd))
-		return 0;
-	if (pmd_devmap(*pmd))
-		return PMD_SHIFT;
-	pte = pte_offset_map(pmd, address);
-	if (!pte_present(*pte))
-		return 0;
-	if (pte_devmap(*pte))
-		return PAGE_SHIFT;
-	return 0;
-}
-
 /*
  * Failure handling: if we can't find or can't kill a process there's
  * not much we can do.	We just print a message and ignore otherwise.

@@ -329,7 +295,9 @@ static void add_to_kill(struct task_struct *tsk, struct page *p,
 	tk->addr = page_address_in_vma(p, vma);
 	tk->addr_valid = 1;
 	if (is_zone_device_page(p))
-		tk->size_shift = dev_pagemap_mapping_shift(p, vma);
+		tk->size_shift =
+			dev_pagemap_mapping_shift(vma_address(page, vma),
+						  vma->vm_mm);
 	else
 		tk->size_shift = compound_order(compound_head(p)) + PAGE_SHIFT;
 

mm/util.c +24 -0

index 8bf08b5b5760..61bc9bab931d 100644
@@ -780,3 +780,37 @@ int get_cmdline(struct task_struct *task, char *buffer, int buflen)
 out:
 	return res;
 }
+
+unsigned long dev_pagemap_mapping_shift(unsigned long address,
+					struct mm_struct *mm)
+{
+	pgd_t *pgd;
+	p4d_t *p4d;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+
+	pgd = pgd_offset(mm, address);
+	if (!pgd_present(*pgd))
+		return 0;
+	p4d = p4d_offset(pgd, address);
+	if (!p4d_present(*p4d))
+		return 0;
+	pud = pud_offset(p4d, address);
+	if (!pud_present(*pud))
+		return 0;
+	if (pud_devmap(*pud))
+		return PUD_SHIFT;
+	pmd = pmd_offset(pud, address);
+	if (!pmd_present(*pmd))
+		return 0;

This change allows KVM to map DAX-backed files made of huge pages with huge mappings in the EPT/TDP.

DAX pages are not PageTransCompound. The existing check is trying to determine if the mapping for the pfn is a huge mapping or not. For non-DAX maps, e.g. hugetlbfs, that means checking PageTransCompound. For DAX, we can check the page table itself.

Note that KVM already faulted in the page (or huge page) in the host's page table, and we hold the KVM mmu spinlock (grabbed before checking the mmu seq).

Signed-off-by Barret Rhoden <[email protected]>

---
 arch/x86/kvm/mmu.c | 34 ++++++++++++++++++++++++++++++++--
 1 file changed, 32 insertions(+), 2 deletions(-)

arch/x86/kvm/mmu.c +32 -2

index cf5f572f2305..2df8c459dc6a 100644
@@ -3152,6 +3152,36 @@ static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, kvm_pfn_t pfn)
 	return -EFAULT;
 }
 
+static bool pfn_is_huge_mapped(struct kvm *kvm, gfn_t gfn, kvm_pfn_t pfn)
+{
+	struct page *page = pfn_to_page(pfn);
+	unsigned long hva, map_shift;
+
+	if (!is_zone_device_page(page))
+		return PageTransCompoundMap(page);
+
+	/*
+	 * DAX pages do not use compound pages.  The page should have already
+	 * been mapped into the host-side page table during try_async_pf(), so
+	 * we can check the page tables directly.
+	 */
+	hva = gfn_to_hva(kvm, gfn);
+	if (kvm_is_error_hva(hva))
+		return false;
+
+	/*
+	 * Our caller grabbed the KVM mmu_lock with a successful
+	 * mmu_notifier_retry, so we're safe to walk the page table.
+	 */
+	map_shift = dev_pagemap_mapping_shift(hva, current->mm);
+	switch (map_shift) {
+	case PMD_SHIFT:
+	case PUD_SIZE:
+		return true;
+	}
+	return false;
+}
+
 static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 					gfn_t *gfnp, kvm_pfn_t *pfnp,
 					int *levelp)

@@ -3168,7 +3198,7 @@ static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 	 */
 	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
 	    level == PT_PAGE_TABLE_LEVEL &&
-	    PageTransCompoundMap(pfn_to_page(pfn)) &&
+	    pfn_is_huge_mapped(vcpu->kvm, gfn, pfn) &&
 	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
 		unsigned long mask;
 		/*

@@ -5678,7 +5708,7 @@ static bool kvm_mmu_zap_collapsible_spte(struct kvm *kvm,
 		 */
 		if (sp->role.direct &&
 			!kvm_is_reserved_pfn(pfn) &&
-			PageTransCompoundMap(pfn_to_page(pfn))) {
+			pfn_is_huge_mapped(kvm, sp->gfn, pfn)) {
 			pte_list_remove(rmap_head, sptep);
 			need_tlb_flush = 1;
 			goto restart;

Looks good. What's the plan for removing PageReserved from DAX pages?

Quoted 1 line · On 2018-11-12 at 20:31 Paolo Bonzini <[email protected]> wrote:

Looks good. What's the plan for removing PageReserved from DAX pages?

I hear that's going on in this thread:

https://lore.kernel.org/lkml/154145268025.30046.11742652345962594283.stgit@ahduyck-desk1.jf.intel.com/

Though it looks like it's speeding up page initialization, and not explicitly making the PageReserved change, yet.

Alternatively, I could change kvm_is_reserved_pfn() to single out zone_device pages if we don't want to wait or if there is a concern that it won't happen.

On a related note, there are two places in KVM where we check PageReserved outside of kvm_is_reserved_pfn().

For reference:

bool kvm_is_reserved_pfn(kvm_pfn_t pfn)

{
        if (pfn_valid(pfn))
                return PageReserved(pfn_to_page(pfn));

        return true;
}

One caller of PageReserved():

void kvm_set_pfn_dirty(kvm_pfn_t pfn)

{
        if (!kvm_is_reserved_pfn(pfn)) {
                struct page *page = pfn_to_page(pfn);

                if (!PageReserved(page))
                        SetPageDirty(page);
        }
}

In that one, the PageReserved() check looks redundant, since if the page was PageReserved, then it would have been kvm_is_reserved.

The other is:

static bool kvm_is_mmio_pfn(kvm_pfn_t pfn)
{
if (pfn_valid(pfn))
return !is_zero_pfn(pfn) && PageReserved(pfn_to_page(pfn)) &&
/*
* Some reserved pages, such as those from NVDIMM
* DAX devices, are not for MMIO, and can be mapped
* with cached memory type for better performance.
* However, the above check misconceives those pages
* as MMIO, and results in KVM mapping them with UC
* memory type, which would hurt the performance.
* Therefore, we check the host memory type in addition
* and only treat UC/UC-/WC pages as MMIO. */
(!pat_enabled() || pat_pfn_immune_to_uc_mtrr(pfn));

return true;
}

Where the PAT stuff was motivated by DAX. The PageReserved check here looks like a broken-out version of kvm_is_reserved_pfn(), so that we can make some extra checks around it.

This change allows KVM to map DAX-backed files made of huge pages with huge mappings in the EPT/TDP.

DAX pages are not PageTransCompound. The existing check is trying to determine if the mapping for the pfn is a huge mapping or not. For non-DAX maps, e.g. hugetlbfs, that means checking PageTransCompound. For DAX, we can check the page table itself.

Note that KVM already faulted in the page (or huge page) in the host's page table, and we hold the KVM mmu spinlock (grabbed before checking the mmu seq).

Signed-off-by Barret Rhoden <[email protected]>

---
 arch/x86/kvm/mmu.c | 34 ++++++++++++++++++++++++++++++++--
 1 file changed, 32 insertions(+), 2 deletions(-)

arch/x86/kvm/mmu.c +12 -0

index cf5f572f2305..2df8c459dc6a 100644
@@ -3152,6 +3152,36 @@ static int kvm_handle_bad_page(struct kvm_vcpu *vcpu, gfn_t gfn, kvm_pfn_t pfn)
 	return -EFAULT;
 }
 
+static bool pfn_is_huge_mapped(struct kvm *kvm, gfn_t gfn, kvm_pfn_t pfn)
+{
+	struct page *page = pfn_to_page(pfn);
+	unsigned long hva, map_shift;
+
+	if (!is_zone_device_page(page))
+		return PageTransCompoundMap(page);
+
+	/*
+	 * DAX pages do not use compound pages.  The page should have already
+	 * been mapped into the host-side page table during try_async_pf(), so
+	 * we can check the page tables directly.

arch/x86/kvm/mmu.c +9 -1

+	switch (map_shift) {
+	case PMD_SHIFT:
+	case PUD_SIZE:
+		return true;
+	}
+	return false;
+}
+
 static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 					gfn_t *gfnp, kvm_pfn_t *pfnp,
 					int *levelp)

@@ -3168,7 +3198,7 @@ static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 	 */
 	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
 	    level == PT_PAGE_TABLE_LEVEL &&
-	    PageTransCompoundMap(pfn_to_page(pfn)) &&
+	    pfn_is_huge_mapped(vcpu->kvm, gfn, pfn) &&
 	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
 		unsigned long mask;
 		/*

Quoted 8 lines

This change allows KVM to map DAX-backed files made of huge pages with huge mappings in the EPT/TDP.

DAX pages are not PageTransCompound. The existing check is trying to determine if the mapping for the pfn is a huge mapping or not. For non-DAX maps, e.g. hugetlbfs, that means checking PageTransCompound. For DAX, we can check the page table itself.

Quoted 3 lines

Note that KVM already faulted in the page (or huge page) in the host's page table, and we hold the KVM mmu spinlock (grabbed before checking the mmu seq).

I wonder if the KVM mmu spinlock is enough for walking (not KVM exclusive) host page tables. Can you elaborate?

Quoted 28 lines

Signed-off-by Barret Rhoden <[email protected]>

---
 arch/x86/kvm/mmu.c | 34 ++++++++++++++++++++++++++++++++--
 1 file changed, 32 insertions(+), 2 deletions(-)

arch/x86/kvm/mmu.c +12 -0

index cf5f572f2305..2df8c459dc6a 100644
@@ -3152,6 +3152,36 @@ static int kvm_handle_bad_page(struct kvm_vcpu
 *vcpu, gfn_t gfn, kvm_pfn_t pfn)
 	return -EFAULT;
 }
 
+static bool pfn_is_huge_mapped(struct kvm *kvm, gfn_t gfn, kvm_pfn_t pfn)
+{
+	struct page *page = pfn_to_page(pfn);
+	unsigned long hva, map_shift;
+
+	if (!is_zone_device_page(page))
+		return PageTransCompoundMap(page);
+
+	/*
+	 * DAX pages do not use compound pages.  The page should have already
+	 * been mapped into the host-side page table during try_async_pf(), so
+	 * we can check the page tables directly.

arch/x86/kvm/mmu.c static int kvm_handle_bad_page(struct kvm_vcpu · line 3177

Quoted 10 lines

arch/x86/kvm/mmu.c +10 -0

+	 */
+	hva = gfn_to_hva(kvm, gfn);
+	if (kvm_is_error_hva(hva))
+		return false;
+
+	/*
+	 * Our caller grabbed the KVM mmu_lock with a successful
+	 * mmu_notifier_retry, so we're safe to walk the page table.
+	 */
+	map_shift = dev_pagemap_mapping_shift(hva, current->mm);

You could get rid of that local variable map_shift.

Quoted 30 lines

arch/x86/kvm/mmu.c +8 -0

+	switch (map_shift) {
+	case PMD_SHIFT:
+	case PUD_SIZE:
+		return true;
+	}
+	return false;
+}
+
 static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 					gfn_t *gfnp, kvm_pfn_t *pfnp,

arch/x86/kvm/mmu.c +1 -1

 					int *levelp)

@@ -3168,7 +3198,7 @@ static void transparent_hugepage_adjust(struct
 kvm_vcpu *vcpu,
 	 */
 	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
 	    level == PT_PAGE_TABLE_LEVEL &&
-	    PageTransCompoundMap(pfn_to_page(pfn)) &&
+	    pfn_is_huge_mapped(vcpu->kvm, gfn, pfn) &&
 	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
 		unsigned long mask;

arch/x86/kvm/mmu.c +1 -1

 		/*

@@ -5678,7 +5708,7 @@ static bool kvm_mmu_zap_collapsible_spte(struct kvm
 *kvm,
 		 */
 		if (sp->role.direct &&
 			!kvm_is_reserved_pfn(pfn) &&
-			PageTransCompoundMap(pfn_to_page(pfn))) {
+			pfn_is_huge_mapped(kvm, sp->gfn, pfn)) {
 			pte_list_remove(rmap_head, sptep);
 			need_tlb_flush = 1;

Quoted 2 lines

goto restart;

This looks surprisingly simple to me :)

--

Thanks,

David / dhildenb

Quoted 15 lines · Barret Rhoden

Quoted 11 lines

This change allows KVM to map DAX-backed files made of huge pages with huge mappings in the EPT/TDP.

DAX pages are not PageTransCompound. The existing check is trying to determine if the mapping for the pfn is a huge mapping or not. For non-DAX maps, e.g. hugetlbfs, that means checking PageTransCompound. For DAX, we can check the page table itself.

Note that KVM already faulted in the page (or huge page) in the host's page table, and we hold the KVM mmu spinlock (grabbed before checking the mmu seq).

Quoted 2 lines

I wonder if the KVM mmu spinlock is enough for walking (not KVM exclusive) host page tables. Can you elaborate?

As this patch is dependent on PageReserved patch(which is in progress), just wondering if we are able to test the code path for hugepage with DAX.

Thanks,
Pankaj

Quoted 82 lines

Quoted 28 lines

Signed-off-by Barret Rhoden <[email protected]>

---
 arch/x86/kvm/mmu.c | 34 ++++++++++++++++++++++++++++++++--
 1 file changed, 32 insertions(+), 2 deletions(-)

arch/x86/kvm/mmu.c +12 -0

index cf5f572f2305..2df8c459dc6a 100644
@@ -3152,6 +3152,36 @@ static int kvm_handle_bad_page(struct kvm_vcpu
 *vcpu, gfn_t gfn, kvm_pfn_t pfn)
 	return -EFAULT;
 }
 
+static bool pfn_is_huge_mapped(struct kvm *kvm, gfn_t gfn, kvm_pfn_t pfn)
+{
+	struct page *page = pfn_to_page(pfn);
+	unsigned long hva, map_shift;
+
+	if (!is_zone_device_page(page))
+		return PageTransCompoundMap(page);
+
+	/*
+	 * DAX pages do not use compound pages.  The page should have already
+	 * been mapped into the host-side page table during try_async_pf(), so
+	 * we can check the page tables directly.

arch/x86/kvm/mmu.c static int kvm_handle_bad_page(struct kvm_vcpu · line 3177

Quoted 10 lines

arch/x86/kvm/mmu.c +10 -0

+	 */
+	hva = gfn_to_hva(kvm, gfn);
+	if (kvm_is_error_hva(hva))
+		return false;
+
+	/*
+	 * Our caller grabbed the KVM mmu_lock with a successful
+	 * mmu_notifier_retry, so we're safe to walk the page table.
+	 */
+	map_shift = dev_pagemap_mapping_shift(hva, current->mm);

You could get rid of that local variable map_shift.

Quoted 30 lines

arch/x86/kvm/mmu.c +8 -0

+	switch (map_shift) {
+	case PMD_SHIFT:
+	case PUD_SIZE:
+		return true;
+	}
+	return false;
+}
+
 static void transparent_hugepage_adjust(struct kvm_vcpu *vcpu,
 					gfn_t *gfnp, kvm_pfn_t *pfnp,

arch/x86/kvm/mmu.c +1 -1

 					int *levelp)

@@ -3168,7 +3198,7 @@ static void transparent_hugepage_adjust(struct
 kvm_vcpu *vcpu,
 	 */
 	if (!is_error_noslot_pfn(pfn) && !kvm_is_reserved_pfn(pfn) &&
 	    level == PT_PAGE_TABLE_LEVEL &&
-	    PageTransCompoundMap(pfn_to_page(pfn)) &&
+	    pfn_is_huge_mapped(vcpu->kvm, gfn, pfn) &&
 	    !mmu_gfn_lpage_is_disallowed(vcpu, gfn, PT_DIRECTORY_LEVEL)) {
 		unsigned long mask;

arch/x86/kvm/mmu.c +1 -1

 		/*

@@ -5678,7 +5708,7 @@ static bool kvm_mmu_zap_collapsible_spte(struct kvm
 *kvm,
 		 */
 		if (sp->role.direct &&
 			!kvm_is_reserved_pfn(pfn) &&
-			PageTransCompoundMap(pfn_to_page(pfn))) {
+			pfn_is_huge_mapped(kvm, sp->gfn, pfn)) {
 			pte_list_remove(rmap_head, sptep);
 			need_tlb_flush = 1;

Quoted 2 lines

goto restart;

This looks surprisingly simple to me :)

--

Thanks,

David / dhildenb

As this patch is dependent on PageReserved patch(which is in progress), just wondering if we are able to test the code path for hugepage with DAX.

Quoted 4 lines

As this patch is dependent on PageReserved patch(which is in progress), just
wondering if we are able to test the code path for hugepage with DAX.

For testing, I used the following patch. It's not 100%, since it intercepts at kvm_is_reserved_pfn(), and not PageReserved() directly. The only difference is with kvm_set_pfn_dirty() I think.

I also have a nasty module that would dump the EPT's and the host page table's mappings so I could confirm that the huge pages are being mapped correctly.

-----------------------

virt/kvm/kvm_main.c +4 -1

index 2679e476b6c3..1b394a0752a0 100644
@@ -148,6 +148,10 @@ __weak int kvm_arch_mmu_notifier_invalidate_range(struct
 kvm *kvm,
 
 bool kvm_is_reserved_pfn(kvm_pfn_t pfn)
 {
+	// XXX hack
+	if (is_zone_device_page(pfn_to_page(pfn)))
+		return false;
+
 	if (pfn_valid(pfn))
 		return PageReserved(pfn_to_page(pfn));
 
-----------------------

Thanks,

Barret

Quoted 3 lines

Note that KVM already faulted in the page (or huge page) in the host's page table, and we hold the KVM mmu spinlock (grabbed before checking the mmu seq).

I wonder if the KVM mmu spinlock is enough for walking (not KVM exclusive) host page tables. Can you elaborate?

KVM has a use case for determining the size of a dax mapping. The KVM code has easy access to the address and the mm; hence the change in parameters.

Signed-off-by Barret Rhoden <[email protected]>

Reviewed-by David Hildenbrand <[email protected]>

---
 include/linux/mm.h  |  3 +++
 mm/memory-failure.c | 38 +++-----------------------------------
 mm/util.c           | 34 ++++++++++++++++++++++++++++++++++
 3 files changed, 40 insertions(+), 35 deletions(-)

include/linux/mm.h +3 -0

index 5411de93a363..51215d695753 100644
@@ -935,6 +935,9 @@ static inline bool is_pci_p2pdma_page(const struct page *page)
 }
 #endif /* CONFIG_DEV_PAGEMAP_OPS */
 
+unsigned long dev_pagemap_mapping_shift(unsigned long address,
+					struct mm_struct *mm);
+
 static inline void get_page(struct page *page)
 {
 	page = compound_head(page);

mm/memory-failure.c +3 -35

index 0cd3de3550f0..c3f2c6a8607e 100644
@@ -265,40 +265,6 @@ void shake_page(struct page *p, int access)
 }
 EXPORT_SYMBOL_GPL(shake_page);
 
-static unsigned long dev_pagemap_mapping_shift(struct page *page,
-		struct vm_area_struct *vma)
-{
-	unsigned long address = vma_address(page, vma);
-	pgd_t *pgd;
-	p4d_t *p4d;
-	pud_t *pud;
-	pmd_t *pmd;
-	pte_t *pte;
-
-	pgd = pgd_offset(vma->vm_mm, address);
-	if (!pgd_present(*pgd))
-		return 0;
-	p4d = p4d_offset(pgd, address);
-	if (!p4d_present(*p4d))
-		return 0;
-	pud = pud_offset(p4d, address);
-	if (!pud_present(*pud))
-		return 0;
-	if (pud_devmap(*pud))
-		return PUD_SHIFT;
-	pmd = pmd_offset(pud, address);
-	if (!pmd_present(*pmd))
-		return 0;
-	if (pmd_devmap(*pmd))
-		return PMD_SHIFT;
-	pte = pte_offset_map(pmd, address);
-	if (!pte_present(*pte))
-		return 0;
-	if (pte_devmap(*pte))
-		return PAGE_SHIFT;
-	return 0;
-}
-
 /*
  * Failure handling: if we can't find or can't kill a process there's
  * not much we can do.	We just print a message and ignore otherwise.

@@ -329,7 +295,9 @@ static void add_to_kill(struct task_struct *tsk, struct page *p,
 	tk->addr = page_address_in_vma(p, vma);
 	tk->addr_valid = 1;
 	if (is_zone_device_page(p))
-		tk->size_shift = dev_pagemap_mapping_shift(p, vma);
+		tk->size_shift =
+			dev_pagemap_mapping_shift(vma_address(page, vma),
+						  vma->vm_mm);
 	else
 		tk->size_shift = compound_order(compound_head(p)) + PAGE_SHIFT;
 

mm/util.c +24 -0

index 8bf08b5b5760..61bc9bab931d 100644
@@ -780,3 +780,37 @@ int get_cmdline(struct task_struct *task, char *buffer, int buflen)
 out:
 	return res;
 }
+
+unsigned long dev_pagemap_mapping_shift(unsigned long address,
+					struct mm_struct *mm)
+{
+	pgd_t *pgd;
+	p4d_t *p4d;
+	pud_t *pud;
+	pmd_t *pmd;
+	pte_t *pte;
+
+	pgd = pgd_offset(mm, address);
+	if (!pgd_present(*pgd))
+		return 0;
+	p4d = p4d_offset(pgd, address);
+	if (!p4d_present(*p4d))
+		return 0;
+	pud = pud_offset(p4d, address);
+	if (!pud_present(*pud))
+		return 0;
+	if (pud_devmap(*pud))
+		return PUD_SHIFT;
+	pmd = pmd_offset(pud, address);
+	if (!pmd_present(*pmd))
+		return 0;

KVM has a use case for determining the size of a dax mapping. The KVM code has easy access to the address and the mm; hence the change in parameters.

Signed-off-by Barret Rhoden <[email protected]>

Reviewed-by David Hildenbrand <[email protected]>

kvm_is_reserved_pfn() already checks PageReserved().

Signed-off-by Barret Rhoden <[email protected]>

---
 virt/kvm/kvm_main.c | 8 ++------
 1 file changed, 2 insertions(+), 6 deletions(-)

virt/kvm/kvm_main.c +0 -1

index 2679e476b6c3..e0ea6d7dac14 100644
@@ -1700,12 +1700,8 @@ EXPORT_SYMBOL_GPL(kvm_release_pfn_dirty);
 
 void kvm_set_pfn_dirty(kvm_pfn_t pfn)
 {
-	if (!kvm_is_reserved_pfn(pfn)) {

This patch series depends on dax pages not being PageReserved. Once that is in place, these changes will let KVM use huge pages with dax-backed files. Without the PageReserved change, KVM and DAX still work with these patches, simply without huge pages - which is the current situation.

RFC/discussion thread:
https://lore.kernel.org/lkml/[email protected]/

v1 -> v2:
https://lore.kernel.org/lkml/[email protected]/

- Updated Acks/Reviewed-by
- Minor touchups
- Added patch to remove redundant PageReserved() check
- Rebased onto linux-next

Barret Rhoden (3):
mm: make dev_pagemap_mapping_shift() externally visible
kvm: Use huge pages for DAX-backed files
kvm: remove redundant PageReserved() check

 arch/x86/kvm/mmu.c  | 33 +++++++++++++++++++++++++++++++--
 include/linux/mm.h  |  3 +++
 mm/memory-failure.c | 38 +++-----------------------------------
 mm/util.c           | 34 ++++++++++++++++++++++++++++++++++
 virt/kvm/kvm_main.c |  8 ++------
 5 files changed, 73 insertions(+), 43 deletions(-)

--
2.19.1.1215.g8438c0b245-goog

[ add Alex who is looking into removing PageReserved for DAX pages. ]

Quoted 34 lines

This patch series depends on dax pages not being PageReserved. Once that is in place, these changes will let KVM use huge pages with dax-backed files. Without the PageReserved change, KVM and DAX still work with these patches, simply without huge pages - which is the current situation.

RFC/discussion thread:
https://lore.kernel.org/lkml/[email protected]/

v1 -> v2:
https://lore.kernel.org/lkml/[email protected]/

- Updated Acks/Reviewed-by
- Minor touchups
- Added patch to remove redundant PageReserved() check
- Rebased onto linux-next

Barret Rhoden (3):
mm: make dev_pagemap_mapping_shift() externally visible
kvm: Use huge pages for DAX-backed files
kvm: remove redundant PageReserved() check

 arch/x86/kvm/mmu.c  | 33 +++++++++++++++++++++++++++++++--
 include/linux/mm.h  |  3 +++
 mm/memory-failure.c | 38 +++-----------------------------------
 mm/util.c           | 34 ++++++++++++++++++++++++++++++++++
 virt/kvm/kvm_main.c |  8 ++------
 5 files changed, 73 insertions(+), 43 deletions(-)

--
2.19.1.1215.g8438c0b245-goog

Quoted 1 line · On 2018-11-14 at 16:55 Dan Williams <dan.j.williams-ral2JQCrhuEAvxtiuMwx3w@publ…

[ add Alex who is looking into removing PageReserved for DAX pages. ]

Thanks. I can keep my eye out for his patches and repost once that's done.

Alternatively, if you all want to merge this before the PageReserved / DAX changes, then I can repost - just Ack/Review tags. It's harmless with the existing PageReserved behavior.

Thanks,

Barret