[PATCH 0/3] KVM: Integrate rmap and rmap_pde

[PATCH 0/3] KVM: Integrate rmap and rmap_pde

[Takuya Yoshikawa]

This has been already discussed on other threads and the concept itself
is not so controversial.

But since I know that the last patch of this series conflicts with
Paul's recent work, I want to find a way to synchronize with his work
at this point: so made the last one an RFC.

Suggestions?

Takuya Yoshikawa (3):
  KVM: Stop checking rmap to see if slot is being created
  KVM: MMU: Use gfn_to_rmap() instead of directly reading rmap array
  KVM: Push rmap into kvm_arch_memory_slot

 arch/powerpc/include/asm/kvm_host.h |    1 +
 arch/powerpc/kvm/book3s_64_mmu_hv.c |    6 ++--
 arch/powerpc/kvm/book3s_hv_rm_mmu.c |    4 +-
 arch/powerpc/kvm/powerpc.c          |    8 +++++
 arch/x86/include/asm/kvm_host.h     |    2 +-
 arch/x86/kvm/mmu.c                  |    8 ++---
 arch/x86/kvm/mmu_audit.c            |    4 +--
 arch/x86/kvm/x86.c                  |   59 ++++++++++++++++++++---------------
 include/linux/kvm_host.h            |    1 -
 virt/kvm/kvm_main.c                 |   11 +------
 10 files changed, 54 insertions(+), 50 deletions(-)

-- 
1.7.5.4

[PATCH 1/3] KVM: Stop checking rmap to see if slot is being created

[Takuya Yoshikawa]

Instead, check npages consistently.  This helps to make rmap
architecture specific in a later patch.

Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
---
 arch/x86/kvm/x86.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index b6379e5..701dbd4 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -6381,7 +6381,7 @@ int kvm_arch_prepare_memory_region(struct kvm *kvm,
 	 *x86 needs to handle !user_alloc case.
 	 */
 	if (!user_alloc) {
-		if (npages && !old.rmap) {
+		if (npages && !old.npages) {
 			unsigned long userspace_addr;
 
 			userspace_addr = vm_mmap(NULL, 0,
@@ -6409,7 +6409,7 @@ void kvm_arch_commit_memory_region(struct kvm *kvm,
 
 	int nr_mmu_pages = 0, npages = mem->memory_size >> PAGE_SHIFT;
 
-	if (!user_alloc && !old.user_alloc && old.rmap && !npages) {
+	if (!user_alloc && !old.user_alloc && old.npages && !npages) {
 		int ret;
 
 		ret = vm_munmap(old.userspace_addr,
-- 
1.7.5.4

[PATCH 2/3] KVM: MMU: Use gfn_to_rmap() instead of directly reading rmap array

[Takuya Yoshikawa]

This helps to make rmap architecture specific in a later patch.

Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
---
 arch/x86/kvm/mmu.c       |    3 ++-
 arch/x86/kvm/mmu_audit.c |    4 +---
 2 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index a9a2052..ee768bb 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -1181,7 +1181,8 @@ void kvm_mmu_write_protect_pt_masked(struct kvm *kvm,
 	unsigned long *rmapp;
 
 	while (mask) {
-		rmapp = &slot->rmap[gfn_offset + __ffs(mask)];
+		rmapp = __gfn_to_rmap(slot->base_gfn + gfn_offset + __ffs(mask),
+				      PT_PAGE_TABLE_LEVEL, slot);
 		__rmap_write_protect(kvm, rmapp, PT_PAGE_TABLE_LEVEL, false);
 
 		/* clear the first set bit */
diff --git a/arch/x86/kvm/mmu_audit.c b/arch/x86/kvm/mmu_audit.c
index 7d7d0b9..ca403f9 100644
--- a/arch/x86/kvm/mmu_audit.c
+++ b/arch/x86/kvm/mmu_audit.c
@@ -190,7 +190,6 @@ static void check_mappings_rmap(struct kvm *kvm, struct kvm_mmu_page *sp)
 
 static void audit_write_protection(struct kvm *kvm, struct kvm_mmu_page *sp)
 {
-	struct kvm_memory_slot *slot;
 	unsigned long *rmapp;
 	u64 *sptep;
 	struct rmap_iterator iter;
@@ -198,8 +197,7 @@ static void audit_write_protection(struct kvm *kvm, struct kvm_mmu_page *sp)
 	if (sp->role.direct || sp->unsync || sp->role.invalid)
 		return;
 
-	slot = gfn_to_memslot(kvm, sp->gfn);
-	rmapp = &slot->rmap[sp->gfn - slot->base_gfn];
+	rmapp = gfn_to_rmap(kvm, sp->gfn, PT_PAGE_TABLE_LEVEL);
 
 	for (sptep = rmap_get_first(*rmapp, &iter); sptep;
 	     sptep = rmap_get_next(&iter)) {
-- 
1.7.5.4

[RFC PATCH 3/3] KVM: Push rmap into kvm_arch_memory_slot

[Takuya Yoshikawa]

Two reasons:
 - x86 can integrate rmap and rmap_pde and remove heuristics in
   __gfn_to_rmap().
 - Some architectures do not need rmap.

Since rmap is one of the most memory consuming stuff in KVM, ppc'd
better restrict the allocation to Book3S HV.

Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
Cc: Paul Mackerras <paulus@samba.org>
---
 arch/powerpc/include/asm/kvm_host.h |    1 +
 arch/powerpc/kvm/book3s_64_mmu_hv.c |    6 ++--
 arch/powerpc/kvm/book3s_hv_rm_mmu.c |    4 +-
 arch/powerpc/kvm/powerpc.c          |    8 +++++
 arch/x86/include/asm/kvm_host.h     |    2 +-
 arch/x86/kvm/mmu.c                  |    5 +--
 arch/x86/kvm/x86.c                  |   55 ++++++++++++++++++++--------------
 include/linux/kvm_host.h            |    1 -
 virt/kvm/kvm_main.c                 |   11 +------
 9 files changed, 49 insertions(+), 44 deletions(-)

diff --git a/arch/powerpc/include/asm/kvm_host.h b/arch/powerpc/include/asm/kvm_host.h
index 572ad01..a29e091 100644
--- a/arch/powerpc/include/asm/kvm_host.h
+++ b/arch/powerpc/include/asm/kvm_host.h
@@ -221,6 +221,7 @@ struct revmap_entry {
 #define KVMPPC_GOT_PAGE		0x80
 
 struct kvm_arch_memory_slot {
+	unsigned long *rmap;
 };
 
 struct kvm_arch {
diff --git a/arch/powerpc/kvm/book3s_64_mmu_hv.c b/arch/powerpc/kvm/book3s_64_mmu_hv.c
index 3c635c0..d95d113 100644
--- a/arch/powerpc/kvm/book3s_64_mmu_hv.c
+++ b/arch/powerpc/kvm/book3s_64_mmu_hv.c
@@ -705,7 +705,7 @@ int kvmppc_book3s_hv_page_fault(struct kvm_run *run, struct kvm_vcpu *vcpu,
 		goto out_unlock;
 	hpte[0] = (hpte[0] & ~HPTE_V_ABSENT) | HPTE_V_VALID;
 
-	rmap = &memslot->rmap[gfn - memslot->base_gfn];
+	rmap = &memslot->arch.rmap[gfn - memslot->base_gfn];
 	lock_rmap(rmap);
 
 	/* Check if we might have been invalidated; let the guest retry if so */
@@ -788,7 +788,7 @@ static int kvm_handle_hva_range(struct kvm *kvm,
 		for (; gfn < gfn_end; ++gfn) {
 			gfn_t gfn_offset = gfn - memslot->base_gfn;
 
-			ret = handler(kvm, &memslot->rmap[gfn_offset], gfn);
+			ret = handler(kvm, &memslot->arch.rmap[gfn_offset], gfn);
 			retval |= ret;
 		}
 	}
@@ -1036,7 +1036,7 @@ long kvmppc_hv_get_dirty_log(struct kvm *kvm, struct kvm_memory_slot *memslot)
 	unsigned long *rmapp, *map;
 
 	preempt_disable();
-	rmapp = memslot->rmap;
+	rmapp = memslot->arch.rmap;
 	map = memslot->dirty_bitmap;
 	for (i = 0; i < memslot->npages; ++i) {
 		if (kvm_test_clear_dirty(kvm, rmapp))
diff --git a/arch/powerpc/kvm/book3s_hv_rm_mmu.c b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
index 5c70d19..56ac1a5 100644
--- a/arch/powerpc/kvm/book3s_hv_rm_mmu.c
+++ b/arch/powerpc/kvm/book3s_hv_rm_mmu.c
@@ -84,7 +84,7 @@ static void remove_revmap_chain(struct kvm *kvm, long pte_index,
 	if (!memslot || (memslot->flags & KVM_MEMSLOT_INVALID))
 		return;
 
-	rmap = real_vmalloc_addr(&memslot->rmap[gfn - memslot->base_gfn]);
+	rmap = real_vmalloc_addr(&memslot->arch.rmap[gfn - memslot->base_gfn]);
 	lock_rmap(rmap);
 
 	head = *rmap & KVMPPC_RMAP_INDEX;
@@ -180,7 +180,7 @@ long kvmppc_h_enter(struct kvm_vcpu *vcpu, unsigned long flags,
 	if (!slot_is_aligned(memslot, psize))
 		return H_PARAMETER;
 	slot_fn = gfn - memslot->base_gfn;
-	rmap = &memslot->rmap[slot_fn];
+	rmap = &memslot->arch.rmap[slot_fn];
 
 	if (!kvm->arch.using_mmu_notifiers) {
 		physp = kvm->arch.slot_phys[memslot->id];
diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
index 87f4dc8..879b14a 100644
--- a/arch/powerpc/kvm/powerpc.c
+++ b/arch/powerpc/kvm/powerpc.c
@@ -302,10 +302,18 @@ long kvm_arch_dev_ioctl(struct file *filp,
 void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 			   struct kvm_memory_slot *dont)
 {
+	if (!dont || free->arch.rmap != dont->arch.rmap) {
+		vfree(free->arch.rmap);
+		free->arch.rmap = NULL;
+	}
 }
 
 int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
 {
+	slot->arch.rmap = vzalloc(npages * sizeof(*slot->arch.rmap));
+	if (!slot->arch.rmap)
+		return -ENOMEM;
+
 	return 0;
 }
 
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 48e7131..1309e69 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -504,7 +504,7 @@ struct kvm_lpage_info {
 };
 
 struct kvm_arch_memory_slot {
-	unsigned long *rmap_pde[KVM_NR_PAGE_SIZES - 1];
+	unsigned long *rmap[KVM_NR_PAGE_SIZES];
 	struct kvm_lpage_info *lpage_info[KVM_NR_PAGE_SIZES - 1];
 };
 
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index ee768bb..aa9a987 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -970,11 +970,8 @@ static unsigned long *__gfn_to_rmap(gfn_t gfn, int level,
 {
 	unsigned long idx;
 
-	if (likely(level == PT_PAGE_TABLE_LEVEL))
-		return &slot->rmap[gfn - slot->base_gfn];
-
 	idx = gfn_to_index(gfn, slot->base_gfn, level);
-	return &slot->arch.rmap_pde[level - PT_DIRECTORY_LEVEL][idx];
+	return &slot->arch.rmap[level - PT_PAGE_TABLE_LEVEL][idx];
 }
 
 /*
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 701dbd4..f9fc9b6 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -6299,14 +6299,18 @@ void kvm_arch_free_memslot(struct kvm_memory_slot *free,
 {
 	int i;
 
-	for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
-		if (!dont || free->arch.rmap_pde[i] != dont->arch.rmap_pde[i]) {
-			kvm_kvfree(free->arch.rmap_pde[i]);
-			free->arch.rmap_pde[i] = NULL;
+	for (i = 0; i < KVM_NR_PAGE_SIZES; ++i) {
+		if (!dont || free->arch.rmap[i] != dont->arch.rmap[i]) {
+			kvm_kvfree(free->arch.rmap[i]);
+			free->arch.rmap[i] = NULL;
 		}
-		if (!dont || free->arch.lpage_info[i] != dont->arch.lpage_info[i]) {
-			kvm_kvfree(free->arch.lpage_info[i]);
-			free->arch.lpage_info[i] = NULL;
+		if (i == 0)
+			continue;
+
+		if (!dont || free->arch.lpage_info[i - 1] !=
+			     dont->arch.lpage_info[i - 1]) {
+			kvm_kvfree(free->arch.lpage_info[i - 1]);
+			free->arch.lpage_info[i - 1] = NULL;
 		}
 	}
 }
@@ -6315,28 +6319,30 @@ int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
 {
 	int i;
 
-	for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
+	for (i = 0; i < KVM_NR_PAGE_SIZES; ++i) {
 		unsigned long ugfn;
 		int lpages;
-		int level = i + 2;
+		int level = i + 1;
 
 		lpages = gfn_to_index(slot->base_gfn + npages - 1,
 				      slot->base_gfn, level) + 1;
 
-		slot->arch.rmap_pde[i] =
-			kvm_kvzalloc(lpages * sizeof(*slot->arch.rmap_pde[i]));
-		if (!slot->arch.rmap_pde[i])
+		slot->arch.rmap[i] =
+			kvm_kvzalloc(lpages * sizeof(*slot->arch.rmap[i]));
+		if (!slot->arch.rmap[i])
 			goto out_free;
+		if (i == 0)
+			continue;
 
-		slot->arch.lpage_info[i] =
-			kvm_kvzalloc(lpages * sizeof(*slot->arch.lpage_info[i]));
-		if (!slot->arch.lpage_info[i])
+		slot->arch.lpage_info[i - 1] = kvm_kvzalloc(lpages *
+					sizeof(*slot->arch.lpage_info[i - 1]));
+		if (!slot->arch.lpage_info[i - 1])
 			goto out_free;
 
 		if (slot->base_gfn & (KVM_PAGES_PER_HPAGE(level) - 1))
-			slot->arch.lpage_info[i][0].write_count = 1;
+			slot->arch.lpage_info[i - 1][0].write_count = 1;
 		if ((slot->base_gfn + npages) & (KVM_PAGES_PER_HPAGE(level) - 1))
-			slot->arch.lpage_info[i][lpages - 1].write_count = 1;
+			slot->arch.lpage_info[i - 1][lpages - 1].write_count = 1;
 		ugfn = slot->userspace_addr >> PAGE_SHIFT;
 		/*
 		 * If the gfn and userspace address are not aligned wrt each
@@ -6348,18 +6354,21 @@ int kvm_arch_create_memslot(struct kvm_memory_slot *slot, unsigned long npages)
 			unsigned long j;
 
 			for (j = 0; j < lpages; ++j)
-				slot->arch.lpage_info[i][j].write_count = 1;
+				slot->arch.lpage_info[i - 1][j].write_count = 1;
 		}
 	}
 
 	return 0;
 
 out_free:
-	for (i = 0; i < KVM_NR_PAGE_SIZES - 1; ++i) {
-		kvm_kvfree(slot->arch.rmap_pde[i]);
-		kvm_kvfree(slot->arch.lpage_info[i]);
-		slot->arch.rmap_pde[i] = NULL;
-		slot->arch.lpage_info[i] = NULL;
+	for (i = 0; i < KVM_NR_PAGE_SIZES; ++i) {
+		kvm_kvfree(slot->arch.rmap[i]);
+		slot->arch.rmap[i] = NULL;
+		if (i == 0)
+			continue;
+
+		kvm_kvfree(slot->arch.lpage_info[i - 1]);
+		slot->arch.lpage_info[i - 1] = NULL;
 	}
 	return -ENOMEM;
 }
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index dbc65f9..3c16f0f 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -213,7 +213,6 @@ struct kvm_memory_slot {
 	gfn_t base_gfn;
 	unsigned long npages;
 	unsigned long flags;
-	unsigned long *rmap;
 	unsigned long *dirty_bitmap;
 	struct kvm_arch_memory_slot arch;
 	unsigned long userspace_addr;
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index bcf973e..14ec567 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -550,16 +550,12 @@ static void kvm_destroy_dirty_bitmap(struct kvm_memory_slot *memslot)
 static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
 				  struct kvm_memory_slot *dont)
 {
-	if (!dont || free->rmap != dont->rmap)
-		vfree(free->rmap);
-
 	if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
 		kvm_destroy_dirty_bitmap(free);
 
 	kvm_arch_free_memslot(free, dont);
 
 	free->npages = 0;
-	free->rmap = NULL;
 }
 
 void kvm_free_physmem(struct kvm *kvm)
@@ -768,11 +764,7 @@ int __kvm_set_memory_region(struct kvm *kvm,
 	if (npages && !old.npages) {
 		new.user_alloc = user_alloc;
 		new.userspace_addr = mem->userspace_addr;
-#ifndef CONFIG_S390
-		new.rmap = vzalloc(npages * sizeof(*new.rmap));
-		if (!new.rmap)
-			goto out_free;
-#endif /* not defined CONFIG_S390 */
+
 		if (kvm_arch_create_memslot(&new, npages))
 			goto out_free;
 	}
@@ -831,7 +823,6 @@ int __kvm_set_memory_region(struct kvm *kvm,
 
 	/* actual memory is freed via old in kvm_free_physmem_slot below */
 	if (!npages) {
-		new.rmap = NULL;
 		new.dirty_bitmap = NULL;
 		memset(&new.arch, 0, sizeof(new.arch));
 	}
-- 
1.7.5.4

Re: [PATCH 0/3] KVM: Integrate rmap and rmap_pde

[Avi Kivity]

On 08/01/2012 11:59 AM, Takuya Yoshikawa wrote:
> This has been already discussed on other threads and the concept itself
> is not so controversial.

Looks good.

> But since I know that the last patch of this series conflicts with
> Paul's recent work, I want to find a way to synchronize with his work
> at this point: so made the last one an RFC.
> 
> Suggestions?

Since the changes are just renames, it should be easy for Paul to
rebase.  But Paul, if you want to delay the last patch, let us know.

-- 
error compiling committee.c: too many arguments to function

Re: [RFC PATCH 3/3] KVM: Push rmap into kvm_arch_memory_slot

[Paul Mackerras]

On Wed, Aug 01, 2012 at 06:03:28PM +0900, Takuya Yoshikawa wrote:
> Two reasons:
>  - x86 can integrate rmap and rmap_pde and remove heuristics in
>    __gfn_to_rmap().
>  - Some architectures do not need rmap.
> 
> Since rmap is one of the most memory consuming stuff in KVM, ppc'd
> better restrict the allocation to Book3S HV.
> 
> Signed-off-by: Takuya Yoshikawa <yoshikawa.takuya@oss.ntt.co.jp>
> Cc: Paul Mackerras <paulus@samba.org>

Looks like a worthwhile change.  I do have some patches in preparation
that will intersect with this, but I can manage that once these
patches go upstream.

Acked-by: Paul Mackerras <paulus@samba.org>

Re: [PATCH 0/3] KVM: Integrate rmap and rmap_pde

[Avi Kivity]

On 08/01/2012 11:59 AM, Takuya Yoshikawa wrote:
> This has been already discussed on other threads and the concept itself
> is not so controversial.
> 
> But since I know that the last patch of this series conflicts with
> Paul's recent work, I want to find a way to synchronize with his work
> at this point: so made the last one an RFC.
> 

Thanks, applied all.



-- 
error compiling committee.c: too many arguments to function