[PATCH 0/4] KVM: Dirty page logging (aka live migration) fixes

[PATCH 0/4] KVM: Dirty page logging (aka live migration) fixes

[Avi Kivity]

This small patchset plugs a few holes in the kvm dirty page logging 
implementation.  With these fixes (which really want to be in Linux 
2.6.21), one can migrate a running virtual machine from one host to 
another.  The virtual machine continues executing while its memory image 
is transferred, so there is only a minimal pause in service.

The patchset is also available as part of the kvm git tree; see 
http://kvm.qumranet.com/kvmwiki/Code.

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

[PATCH 1/4] KVM: Add missing calls to mark_page_dirty()

[Avi Kivity]

A few places where we modify guest memory fail to call mark_page_dirty(),
causing live migration to fail.  This adds the missing calls.

Signed-off-by: Uri Lublin <uril@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
---
 drivers/kvm/kvm_main.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index a593d09..edff405 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -228,12 +228,15 @@ int kvm_write_guest(struct kvm_vcpu *vcpu, gva_t addr, unsigned long size,
 		unsigned now;
 		unsigned offset;
 		hva_t guest_buf;
+		gfn_t gfn;
 
 		paddr = gva_to_hpa(vcpu, addr);
 
 		if (is_error_hpa(paddr))
 			break;
 
+		gfn = vcpu->mmu.gva_to_gpa(vcpu, addr) >> PAGE_SHIFT;
+		mark_page_dirty(vcpu->kvm, gfn);
 		guest_buf = (hva_t)kmap_atomic(
 				pfn_to_page(paddr >> PAGE_SHIFT), KM_USER0);
 		offset = addr & ~PAGE_MASK;
@@ -953,6 +956,7 @@ static int emulator_write_phys(struct kvm_vcpu *vcpu, gpa_t gpa,
 		return 0;
 	page = gfn_to_page(m, gpa >> PAGE_SHIFT);
 	kvm_mmu_pre_write(vcpu, gpa, bytes);
+	mark_page_dirty(vcpu->kvm, gpa >> PAGE_SHIFT);
 	virt = kmap_atomic(page, KM_USER0);
 	memcpy(virt + offset_in_page(gpa), &val, bytes);
 	kunmap_atomic(virt, KM_USER0);
@@ -1294,6 +1298,7 @@ static int vcpu_register_para(struct kvm_vcpu *vcpu, gpa_t para_state_gpa)
 	if (is_error_hpa(para_state_hpa))
 		goto err_gp;
 
+	mark_page_dirty(vcpu->kvm, para_state_gpa >> PAGE_SHIFT);
 	para_state_page = pfn_to_page(para_state_hpa >> PAGE_SHIFT);
 	para_state = kmap_atomic(para_state_page, KM_USER0);
 
@@ -1323,6 +1328,7 @@ static int vcpu_register_para(struct kvm_vcpu *vcpu, gpa_t para_state_gpa)
 	vcpu->para_state_gpa = para_state_gpa;
 	vcpu->hypercall_gpa = hypercall_gpa;
 
+	mark_page_dirty(vcpu->kvm, hypercall_gpa >> PAGE_SHIFT);
 	hypercall = kmap_atomic(pfn_to_page(hypercall_hpa >> PAGE_SHIFT),
 				KM_USER1) + (hypercall_hpa & ~PAGE_MASK);
 	kvm_arch_ops->patch_hypercall(vcpu, hypercall);

[PATCH 2/4] KVM: Fix dirty page log bitmap size/access calculation

[Avi Kivity]

Since dirty_bitmap is an unsigned long array, the alignment and size need
to take that into account.

Signed-off-by: Uri Lublin <uril@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
---
 drivers/kvm/kvm_main.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index edff405..e710810 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -792,9 +792,9 @@ static int kvm_vm_ioctl_get_dirty_log(struct kvm *kvm,
 	if (!memslot->dirty_bitmap)
 		goto out;
 
-	n = ALIGN(memslot->npages, 8) / 8;
+	n = ALIGN(memslot->npages, BITS_PER_LONG) / 8;
 
-	for (i = 0; !any && i < n; ++i)
+	for (i = 0; !any && i < n/sizeof(long); ++i)
 		any = memslot->dirty_bitmap[i];
 
 	r = -EFAULT;

[PATCH 3/4] kvm: move do_remove_write_access() up

[Avi Kivity]

To be called from kvm_vm_ioctl_set_memory_region()

Signed-off-by: Uri Lublin <uril@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
---
 drivers/kvm/kvm_main.c |   14 +++++++-------
 1 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index e710810..be7694d 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -611,6 +611,13 @@ void fx_init(struct kvm_vcpu *vcpu)
 }
 EXPORT_SYMBOL_GPL(fx_init);
 
+static void do_remove_write_access(struct kvm_vcpu *vcpu, int slot)
+{
+	spin_lock(&vcpu->kvm->lock);
+	kvm_mmu_slot_remove_write_access(vcpu, slot);
+	spin_unlock(&vcpu->kvm->lock);
+}
+
 /*
  * Allocate some memory and give it an address in the guest physical address
  * space.
@@ -756,13 +763,6 @@ out:
 	return r;
 }
 
-static void do_remove_write_access(struct kvm_vcpu *vcpu, int slot)
-{
-	spin_lock(&vcpu->kvm->lock);
-	kvm_mmu_slot_remove_write_access(vcpu, slot);
-	spin_unlock(&vcpu->kvm->lock);
-}
-
 /*
  * Get (and clear) the dirty memory log for a memory slot.
  */

[PATCH 4/4] KVM: Remove write access permissions when dirty-page-logging is enabled

[Avi Kivity]

Enabling dirty page logging is done using KVM_SET_MEMORY_REGION ioctl.
If the memory region already exists, we need to remove write accesses,
so writes will be caught, and dirty pages will be logged.

Signed-off-by: Uri Lublin <uril@qumranet.com>
Signed-off-by: Avi Kivity <avi@qumranet.com>
---
 drivers/kvm/kvm_main.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index be7694d..e48b4d7 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -748,6 +748,8 @@ raced:
 		vcpu = vcpu_load_slot(kvm, i);
 		if (!vcpu)
 			continue;
+		if (new.flags & KVM_MEM_LOG_DIRTY_PAGES)
+			do_remove_write_access(vcpu, mem->slot);
 		kvm_mmu_reset_context(vcpu);
 		vcpu_put(vcpu);
 	}

Re: [PATCH 0/4] KVM: Dirty page logging (aka live migration) fixes

[Avi Kivity]

Avi Kivity wrote:
> This small patchset plugs a few holes in the kvm dirty page logging 
> implementation.  With these fixes (which really want to be in Linux 
> 2.6.21), one can migrate a running virtual machine from one host to 
> another.  The virtual machine continues executing while its memory 
> image is transferred, so there is only a minimal pause in service.
>
> The patchset is also available as part of the kvm git tree; see 
> http://kvm.qumranet.com/kvmwiki/Code.
>

The patches are all

  From: Uri Lublin <uril@qumranet.com>



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