Add new data type, struct kvm_gfn_ring, and support functions for
ring-based dirty memory logging. Add new ioctl, KVM_RESET_DIRTY_PAGES,
for dirty trap reset.
Signed-off-by: Lei Cao <lei.cao@xxxxxxxxxxx>
---
arch/x86/include/asm/kvm_host.h | 2 --
include/linux/kvm_gfn_ring.h | 68 +++++++++++++++++++++++++++++++++++++++++
include/uapi/linux/kvm.h | 35 +++++++++++++++++++++
virt/kvm/kvm_main.c | 16 ++++++++++
4 files changed, 119 insertions(+), 2 deletions(-)
create mode 100644 include/linux/kvm_gfn_ring.h
diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index a7066dc..6646aa5 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -43,8 +43,6 @@
#define KVM_PRIVATE_MEM_SLOTS 3
#define KVM_MEM_SLOTS_NUM (KVM_USER_MEM_SLOTS + KVM_PRIVATE_MEM_SLOTS)
-#define KVM_PIO_PAGE_OFFSET 1
-#define KVM_COALESCED_MMIO_PAGE_OFFSET 2
#define KVM_HALT_POLL_NS_DEFAULT 400000
#define KVM_IRQCHIP_NUM_PINS KVM_IOAPIC_NUM_PINS
diff --git a/include/linux/kvm_gfn_ring.h b/include/linux/kvm_gfn_ring.h
new file mode 100644
index 0000000..9d5ca99
--- /dev/null
+++ b/include/linux/kvm_gfn_ring.h
@@ -0,0 +1,68 @@
+#ifndef KVM_GFN_RING_H
+#define KVM_GFN_RING_H
+
+/*
+ * struct kvm_dirty_ring is defined in include/uapi/linux/kvm.h.
+ *
+ * dirty_ring: shared with userspace via mmap. dirty_ring->dirty_gfns
+ * is the compact list that holds the dirty pages.
+ * dirty_index: free running counter that points to the next slot in
+ * dirty_ring->dirty_gfns where a new dirty page should go.
+ * reset_index: free running counter that points to the next dirty page
+ * in dirty_ring->dirty_gfns for which dirty trap needs to
+ * be reenabled
+ * size: size of the compact list, dirty_ring->dirty_gfns
+ * soft_limit: when the number of dirty pages in the list reaches this
+ * limit, vcpu that owns this ring should exit to userspace
+ * to allow userspace to harvest all the dirty pages
+ * lock: protects dirty_ring, only in use if this is the global
+ * ring
+ *
+ * The number of dirty pages in the ring is calculated by,
+ * dirty_index - reset_index
+ *
+ * kernel increments dirty_ring->indices.avail_index after dirty index
+ * is incremented. When userspace harvests the dirty pages, it increments
+ * dirty_ring->indices.fetch_index up to dirty_ring->indices.avail_index.
+ * When kernel reenables dirty traps for the dirty pages, it increments
+ * reset_index up to dirty_ring->indices.fetch_index.
+ *
+ */
+struct kvm_gfn_ring {
+ u16 dirty_index;
+ u16 reset_index;
+ u32 size;
+ u32 soft_limit;
+ spinlock_t lock;
+ struct kvm_dirty_ring *dirty_ring;
+};
+
+int kvm_gfn_ring_alloc(struct kvm_gfn_ring *gfnring,
+ u32 size,
+ u32 limit);
+
+/*
+ * called with kvm->slots_lock held, returns the number of
+ * processed pages.
+ */
+int kvm_gfn_ring_reset(struct kvm *kvm,
+ struct kvm_gfn_ring *gfnring);
+
+/*
+ * returns 0: successfully pushed
+ * 1: successfully pushed, soft limit reached,
+ * vcpu should exit to userspace
+ * -EBUSY: unable to push, dirty ring full.
+ */
+int kvm_gfn_ring_push(struct kvm_gfn_ring *gfnring,
+ u32 slot,
+ u64 offset,
+ bool locked);
+
+/* for use in vm_operations_struct */
+struct page *kvm_gfn_ring_get_page(struct kvm_gfn_ring *ring,
+ u32 i);
+
+void kvm_gfn_ring_free(struct kvm_gfn_ring *ring);
+
+#endif
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index 117f1f9..1a1ba4d 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -214,6 +214,10 @@ struct kvm_hyperv_exit {
/* Encounter unexpected vm-exit due to delivery event. */
#define KVM_INTERNAL_ERROR_DELIVERY_EV 3
+#define KVM_PIO_PAGE_OFFSET 1
+#define KVM_COALESCED_MMIO_PAGE_OFFSET 2
+#define KVM_DIRTY_LOG_PAGE_OFFSET 64
+
/* for KVM_RUN, returned by mmap(vcpu_fd, offset=0) */
struct kvm_run {
/* in */
@@ -1283,6 +1287,8 @@ struct kvm_s390_ucas_mapping {
/* Available with KVM_CAP_X86_SMM */
#define KVM_SMI _IO(KVMIO, 0xb7)
+#define KVM_RESET_DIRTY_PAGES _IO(KVMIO, 0xe0)
+
#define KVM_DEV_ASSIGN_ENABLE_IOMMU (1 << 0)
#define KVM_DEV_ASSIGN_PCI_2_3 (1 << 1)
#define KVM_DEV_ASSIGN_MASK_INTX (1 << 2)
@@ -1336,4 +1342,33 @@ struct kvm_assigned_msix_entry {
#define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0)
#define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1)
+/*
+ * The following are the requirements for supporting dirty log ring
+ * (by enabling KVM_DIRTY_LOG_PAGE_OFFSET).
+ *
+ * 1. Memory accesses by KVM should call kvm_vcpu_write_* instead
+ * of kvm_write_* so that the global dirty ring is not filled up
+ * too quickly.
+ * 2. kvm_arch_mmu_enable_log_dirty_pt_masked should be defined for
+ * enabling dirty logging.
+ * 3. There should not be a separate step to synchronize hardware
+ * dirty bitmap with KVM's.
+ */
+
+struct kvm_dirty_gfn {
+ __u32 pad;
+ __u32 slot;
+ __u64 offset;
+};
+
+struct kvm_dirty_ring {
+ union {
+ struct {
+ __u16 avail_index; /* set by kernel */
+ __u16 fetch_index; /* set by userspace */
+ } indices;
+ struct kvm_dirty_gfn dirty_gfns[0];
+ };
+};
+
#endif /* __LINUX_KVM_H */
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index f2744ce..016be4d 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2945,11 +2945,18 @@ static long kvm_vm_ioctl_check_extension_generic(struct kvm *kvm, long arg)
return kvm_vm_ioctl_check_extension(kvm, arg);
}
+#ifdef KVM_DIRTY_LOG_PAGE_OFFSET
static int kvm_vm_ioctl_enable_dirty_log_ring(struct kvm *kvm, __u32 size)
{
return -EINVAL;
}
+static int kvm_vm_ioctl_reset_dirty_pages(struct kvm *kvm)
+{
+ return -EINVAL;
+}
+#endif
+
int __attribute__((weak)) kvm_vm_ioctl_enable_cap(struct kvm *kvm,
struct kvm_enable_cap *cap)
{
@@ -2961,7 +2968,11 @@ static int kvm_vm_ioctl_enable_cap_generic(struct kvm *kvm,
{
switch (cap->cap) {
case KVM_CAP_DIRTY_LOG_RING:
+#ifdef KVM_DIRTY_LOG_PAGE_OFFSET
return kvm_vm_ioctl_enable_dirty_log_ring(kvm, cap->args[0]);
+#else
+ break;
+#endif
default:
return kvm_vm_ioctl_enable_cap(kvm, cap);
}
@@ -3135,6 +3146,11 @@ static long kvm_vm_ioctl(struct file *filp,
case KVM_CHECK_EXTENSION:
r = kvm_vm_ioctl_check_extension_generic(kvm, arg);
break;
+#ifdef KVM_DIRTY_LOG_PAGE_OFFSET
+ case KVM_RESET_DIRTY_PAGES:
+ r = kvm_vm_ioctl_reset_dirty_pages(kvm);
+ break;
+#endif /* KVM_DIRTY_LOG_PAGE_OFFSET */
default:
r = kvm_arch_vm_ioctl(filp, ioctl, arg);
}
--
2.5.0
