KVM: Use a shared page for kernel/user communication when runing a vcpu

Instead of passing a 'struct kvm_run' back and forth between the kernel and
userspace, allocate a page and allow the user to mmap() it.  This reduces
needless copying and makes the interface expandable by providing lots of
free space.

Signed-off-by: Avi Kivity <avi@qumranet.com>

3 files changed, 44 insertions, 17 deletions

diff --git a/drivers/kvm/kvm.h b/drivers/kvm/kvm.h
index 0d122bf889db..901b8d917b55 100644
--- a/drivers/kvm/kvm.h
+++ b/drivers/kvm/kvm.h
@@ -228,6 +228,7 @@ struct kvm_vcpu {
         struct mutex mutex;
         int   cpu;
         int   launched;
+        struct kvm_run *run;
         int interrupt_window_open;
         unsigned long irq_summary; /* bit vector: 1 per word in irq_pending */
 #define NR_IRQ_WORDS KVM_IRQ_BITMAP_SIZE(unsigned long)
diff --git a/drivers/kvm/kvm_main.c b/drivers/kvm/kvm_main.c
index 946ed86a0595..42be8a8f299d 100644
--- a/drivers/kvm/kvm_main.c
+++ b/drivers/kvm/kvm_main.c
@@ -355,6 +355,8 @@ static void kvm_free_vcpu(struct kvm_vcpu *vcpu)
         kvm_mmu_destroy(vcpu);
         vcpu_put(vcpu);
         kvm_arch_ops->vcpu_free(vcpu);
+        free_page((unsigned long)vcpu->run);
+        vcpu->run = NULL;
 }
 
 static void kvm_free_vcpus(struct kvm *kvm)
@@ -1887,6 +1889,33 @@ static int kvm_vcpu_ioctl_debug_guest(struct kvm_vcpu *vcpu,
         return r;
 }
 
+static struct page *kvm_vcpu_nopage(struct vm_area_struct *vma,
+                                    unsigned long address,
+                                    int *type)
+{
+        struct kvm_vcpu *vcpu = vma->vm_file->private_data;
+        unsigned long pgoff;
+        struct page *page;
+
+        *type = VM_FAULT_MINOR;
+        pgoff = ((address - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff;
+        if (pgoff != 0)
+                return NOPAGE_SIGBUS;
+        page = virt_to_page(vcpu->run);
+        get_page(page);
+        return page;
+}
+
+static struct vm_operations_struct kvm_vcpu_vm_ops = {
+        .nopage = kvm_vcpu_nopage,
+};
+
+static int kvm_vcpu_mmap(struct file *file, struct vm_area_struct *vma)
+{
+        vma->vm_ops = &kvm_vcpu_vm_ops;
+        return 0;
+}
+
 static int kvm_vcpu_release(struct inode *inode, struct file *filp)
 {
         struct kvm_vcpu *vcpu = filp->private_data;
@@ -1899,6 +1928,7 @@ static struct file_operations kvm_vcpu_fops = {
         .release        = kvm_vcpu_release,
         .unlocked_ioctl = kvm_vcpu_ioctl,
         .compat_ioctl   = kvm_vcpu_ioctl,
+        .mmap           = kvm_vcpu_mmap,
 };
 
 /*
@@ -1947,6 +1977,7 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n)
 {
         int r;
         struct kvm_vcpu *vcpu;
+        struct page *page;
 
         r = -EINVAL;
         if (!valid_vcpu(n))
@@ -1961,6 +1992,12 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n)
                 return -EEXIST;
         }
 
+        page = alloc_page(GFP_KERNEL | __GFP_ZERO);
+        r = -ENOMEM;
+        if (!page)
+                goto out_unlock;
+        vcpu->run = page_address(page);
+
         vcpu->host_fx_image = (char*)ALIGN((hva_t)vcpu->fx_buf,
                                            FX_IMAGE_ALIGN);
         vcpu->guest_fx_image = vcpu->host_fx_image + FX_IMAGE_SIZE;
@@ -1990,6 +2027,7 @@ static int kvm_vm_ioctl_create_vcpu(struct kvm *kvm, int n)
 
 out_free_vcpus:
         kvm_free_vcpu(vcpu);
+out_unlock:
         mutex_unlock(&vcpu->mutex);
 out:
         return r;
@@ -2003,21 +2041,9 @@ static long kvm_vcpu_ioctl(struct file *filp,
         int r = -EINVAL;
 
         switch (ioctl) {
-        case KVM_RUN: {
-                struct kvm_run kvm_run;
-
-                r = -EFAULT;
-                if (copy_from_user(&kvm_run, argp, sizeof kvm_run))
-                        goto out;
-                r = kvm_vcpu_ioctl_run(vcpu, &kvm_run);
-                if (r < 0 &&  r != -EINTR)
-                        goto out;
-                if (copy_to_user(argp, &kvm_run, sizeof kvm_run)) {
-                        r = -EFAULT;
-                        goto out;
-                }
+        case KVM_RUN:
+                r = kvm_vcpu_ioctl_run(vcpu, vcpu->run);
                 break;
-        }
         case KVM_GET_REGS: {
                 struct kvm_regs kvm_regs;
 
diff --git a/include/linux/kvm.h b/include/linux/kvm.h
index 275354ffa1cb..d88e7508ee0a 100644
--- a/include/linux/kvm.h
+++ b/include/linux/kvm.h
@@ -11,7 +11,7 @@
 #include <asm/types.h>
 #include <linux/ioctl.h>
 
-#define KVM_API_VERSION 4
+#define KVM_API_VERSION 5
 
 /*
  * Architectural interrupt line count, and the size of the bitmap needed
@@ -49,7 +49,7 @@ enum kvm_exit_reason {
         KVM_EXIT_SHUTDOWN         = 8,
 };
 
-/* for KVM_RUN */
+/* for KVM_RUN, returned by mmap(vcpu_fd, offset=0) */
 struct kvm_run {
         /* in */
         __u32 emulated;  /* skip current instruction */
@@ -233,7 +233,7 @@ struct kvm_dirty_log {
 /*
  * ioctls for vcpu fds
  */
-#define KVM_RUN                   _IOWR(KVMIO, 2, struct kvm_run)
+#define KVM_RUN                   _IO(KVMIO, 16)
 #define KVM_GET_REGS              _IOR(KVMIO, 3, struct kvm_regs)
 #define KVM_SET_REGS              _IOW(KVMIO, 4, struct kvm_regs)
 #define KVM_GET_SREGS             _IOR(KVMIO, 5, struct kvm_sregs)