Linux Virtualization - Mar 2008 - [RFC/PATCH 15/15] guest: virtio device support, and kvm hypercalls

From: Christian Borntraeger <borntraeger at de.ibm.com>

This patch implements kvm guest kernel support for paravirtualized devices
and contains two parts:
o a basic virtio stub using virtio_ring and external interrupts and hypercalls
o full hypercall implementation in kvm_para.h

Currently we dont have PCI on s390. Making virtio_pci usable for s390 seems
more complicated that providing an own stub. This virtio stub is similar to
the lguest one, the memory for the descriptors and the device detection is made
via additional mapped memory on top of the guest storage. We use an external
interrupt with extint code 1237 for host->guest notification. 

The hypercall definition uses the diag instruction for issuing a hypercall. The
parameters are written in R2-R7, the hypercall number is written in R1. This is
similar to the system call ABI (svc) which can use R1 for the number and R2-R6 
for the parameters.


Signed-off-by: Christian Borntraeger <borntraeger at de.ibm.com>
Acked-by: Martin Schwidefsky <schwidefsky at de.ibm.com>
Signed-off-by: Carsten Otte <cotte at de.ibm.com>
---
 drivers/s390/Makefile         |    2 
 drivers/s390/kvm/Makefile     |    9 +
 drivers/s390/kvm/kvm_virtio.c |  326 ++++++++++++++++++++++++++++++++++++++++++
 drivers/s390/kvm/kvm_virtio.h |   47 ++++++
 include/asm-s390/kvm_para.h   |  124 +++++++++++++++
 5 files changed, 505 insertions(+), 3 deletions(-)

Index: kvm/drivers/s390/Makefile
==================================================================---
kvm.orig/drivers/s390/Makefile
+++ kvm/drivers/s390/Makefile
@@ -5,7 +5,7 @@
 CFLAGS_sysinfo.o += -Iinclude/math-emu -Iarch/s390/math-emu -w
 
 obj-y += s390mach.o sysinfo.o s390_rdev.o
-obj-y += cio/ block/ char/ crypto/ net/ scsi/
+obj-y += cio/ block/ char/ crypto/ net/ scsi/ kvm/
 
 drivers-y += drivers/s390/built-in.o
 
Index: kvm/drivers/s390/kvm/Makefile
==================================================================--- /dev/null
+++ kvm/drivers/s390/kvm/Makefile
@@ -0,0 +1,9 @@
+# Makefile for kvm guest drivers on s390
+#
+# Copyright IBM Corp. 2008
+#
+# This program is free software; you can redistribute it and/or modify
+# it under the terms of the GNU General Public License (version 2 only)
+# as published by the Free Software Foundation.
+
+obj-$(CONFIG_VIRTIO) += kvm_virtio.o
Index: kvm/drivers/s390/kvm/kvm_virtio.c
==================================================================--- /dev/null
+++ kvm/drivers/s390/kvm/kvm_virtio.c
@@ -0,0 +1,326 @@
+/*
+ * kvm_virtio.c - virtio for kvm on s390
+ *
+ * Copyright IBM Corp. 2008
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License (version 2 only)
+ * as published by the Free Software Foundation.
+ *
+ *    Author(s): Christian Borntraeger <borntraeger at de.ibm.com>
+ */
+
+#include <linux/init.h>
+#include <linux/bootmem.h>
+#include <linux/err.h>
+#include <linux/virtio.h>
+#include <linux/virtio_config.h>
+#include <linux/interrupt.h>
+#include <linux/virtio_ring.h>
+#include <asm/io.h>
+#include <asm/kvm_para.h>
+#include <asm/setup.h>
+#include <asm/s390_ext.h>
+
+#include "kvm_virtio.h"
+
+/*
+ * The pointer to our (page) of device descriptions.
+ */
+static void *kvm_devices;
+
+/*
+ * Unique numbering for kvm devices.
+ */
+static unsigned int dev_index;
+
+struct kvm_device {
+	struct virtio_device vdev;
+	struct kvm_device_desc *desc;
+};
+
+#define to_kvmdev(vd) container_of(vd, struct kvm_device, vdev)
+
+/*
+ * memory layout:
+ * - kvm_device_descriptor
+ *        struct kvm_device_desc
+ * - configuration
+ *        struct kvm_vqconfig
+ * - feature bits
+ * - config space
+ */
+static struct kvm_vqconfig *kvm_vq_config(const struct kvm_device_desc *desc)
+{
+	return (struct kvm_vqconfig *)(desc + 1);
+}
+
+static u8 *kvm_vq_features(const struct kvm_device_desc *desc)
+{
+	return (u8 *)(kvm_vq_config(desc) + desc->num_vq);
+}
+
+static u8 *kvm_vq_configspace(const struct kvm_device_desc *desc)
+{
+	return kvm_vq_features(desc) + desc->feature_len * 2;
+}
+
+/*
+ * The total size of the config page used by this device (incl. desc)
+ */
+static unsigned desc_size(const struct kvm_device_desc *desc)
+{
+	return sizeof(*desc)
+		+ desc->num_vq * sizeof(struct kvm_vqconfig)
+		+ desc->feature_len * 2
+		+ desc->config_len;
+}
+
+/*
+ * This tests (and acknowleges) a feature bit.
+ */
+static bool kvm_feature(struct virtio_device *vdev, unsigned fbit)
+{
+	struct kvm_device_desc *desc = to_kvmdev(vdev)->desc;
+	u8 *features;
+
+	if (fbit / 8 > desc->feature_len)
+		return false;
+
+	features = kvm_vq_features(desc);
+	if (!(features[fbit / 8] & (1 << (fbit % 8))))
+		return false;
+
+	/*
+	 * We set the matching bit in the other half of the bitmap to tell the
+	 * Host we want to use this feature.
+	 */
+	features[desc->feature_len + fbit / 8] |= (1 << (fbit % 8));
+	return true;
+}
+
+/*
+ * Reading and writing elements in config space
+ */
+static void kvm_get(struct virtio_device *vdev, unsigned int offset,
+		   void *buf, unsigned len)
+{
+	struct kvm_device_desc *desc = to_kvmdev(vdev)->desc;
+
+	BUG_ON(offset + len > desc->config_len);
+	memcpy(buf, kvm_vq_configspace(desc) + offset, len);
+}
+
+static void kvm_set(struct virtio_device *vdev, unsigned int offset,
+		   const void *buf, unsigned len)
+{
+	struct kvm_device_desc *desc = to_kvmdev(vdev)->desc;
+
+	BUG_ON(offset + len > desc->config_len);
+	memcpy(kvm_vq_configspace(desc) + offset, buf, len);
+}
+
+/*
+ * The operations to get and set the status word just access
+ * the status field of the device descriptor.
+ */
+static u8 kvm_get_status(struct virtio_device *vdev)
+{
+	return to_kvmdev(vdev)->desc->status;
+}
+
+static void kvm_set_status(struct virtio_device *vdev, u8 status)
+{
+	BUG_ON(!status);
+	to_kvmdev(vdev)->desc->status = status;
+}
+
+/*
+ * To reset the device, we (ab)use the NOTIFY hypercall, with the descriptor
+ * address of the device.  The Host will zero the status and all the
+ * features.
+ */
+static void kvm_reset(struct virtio_device *vdev)
+{
+	unsigned long offset = (void *)to_kvmdev(vdev)->desc - kvm_devices;
+
+	kvm_hypercall1(1237, (max_pfn<<PAGE_SHIFT) + offset);
+}
+
+/*
+ * When the virtio_ring code wants to notify the Host, it calls us here and we
+ * make a hypercall.  We hand the address  of the virtqueue so the Host
+ * knows which virtqueue we're talking about.
+ */
+static void kvm_notify(struct virtqueue *vq)
+{
+	struct kvm_vqconfig *config = vq->priv;
+
+	kvm_hypercall1(1237, config->address);
+}
+
+/*
+ * This routine finds the first virtqueue described in the configuration of
+ * this device and sets it up.
+ */
+static struct virtqueue *kvm_find_vq(struct virtio_device *vdev,
+				    unsigned index,
+				    void (*callback)(struct virtqueue *vq))
+{
+	struct kvm_device *kdev = to_kvmdev(vdev);
+	struct kvm_vqconfig *config;
+	struct virtqueue *vq;
+	int err;
+
+	if (index >= kdev->desc->num_vq)
+		return ERR_PTR(-ENOENT);
+
+	config = kvm_vq_config(kdev->desc)+index;
+
+	if (add_shared_memory(config->address,
+				vring_size(config->num, PAGE_SIZE))) {
+		err = -ENOMEM;
+		goto out;
+	}
+
+	vq = vring_new_virtqueue(config->num, vdev, (void *) config->address,
+				 kvm_notify, callback);
+	if (!vq) {
+		err = -ENOMEM;
+		goto unmap;
+	}
+
+	/*
+	 * register a callback token
+	 * The host will sent this via the external interrupt parameter
+	 */
+	config->token = (u64) vq;
+
+	vq->priv = config;
+	return vq;
+unmap:
+	remove_shared_memory(config->address, vring_size(config->num,
+			     PAGE_SIZE));
+out:
+	return ERR_PTR(err);
+}
+
+static void kvm_del_vq(struct virtqueue *vq)
+{
+	struct kvm_vqconfig *config = vq->priv;
+
+	vring_del_virtqueue(vq);
+	remove_shared_memory(config->address,
+			     vring_size(config->num, PAGE_SIZE));
+}
+
+/*
+ * The config ops structure as defined by virtio config
+ */
+static struct virtio_config_ops kvm_vq_configspace_ops = {
+	.feature = kvm_feature,
+	.get = kvm_get,
+	.set = kvm_set,
+	.get_status = kvm_get_status,
+	.set_status = kvm_set_status,
+	.reset = kvm_reset,
+	.find_vq = kvm_find_vq,
+	.del_vq = kvm_del_vq,
+};
+
+/*
+ * The root device for the kvm virtio devices.
+ * This makes them appear as /sys/devices/kvm/0,1,2 not /sys/devices/0,1,2.
+ */
+static struct device kvm_root = {
+	.parent = NULL,
+	.bus_id = "kvm_s390",
+};
+
+/*
+ * adds a new device and register it with virtio
+ * appropriate drivers are loaded by the device model
+ */
+static void add_kvm_device(struct kvm_device_desc *d)
+{
+	struct kvm_device *kdev;
+
+	kdev = kzalloc(sizeof(*kdev), GFP_KERNEL);
+	if (!kdev) {
+		printk(KERN_EMERG "Cannot allocate kvm dev %u\n",
+		       dev_index++);
+		return;
+	}
+
+	kdev->vdev.dev.parent = &kvm_root;
+	kdev->vdev.index = dev_index++;
+	kdev->vdev.id.device = d->type;
+	kdev->vdev.config = &kvm_vq_configspace_ops;
+	kdev->desc = d;
+
+	if (register_virtio_device(&kdev->vdev) != 0) {
+		printk(KERN_ERR "Failed to register kvm device %u\n",
+		       kdev->vdev.index);
+		kfree(kdev);
+	}
+}
+
+/*
+ * scan_devices() simply iterates through the device page.
+ * The type 0 is reserved to mean "end of devices".
+ */
+static void scan_devices(void)
+{
+	unsigned int i;
+	struct kvm_device_desc *d;
+
+	for (i = 0; i < PAGE_SIZE; i += desc_size(d)) {
+		d = kvm_devices + i;
+
+		if (d->type == 0)
+			break;
+
+		add_kvm_device(d);
+	}
+}
+
+/*
+ * we emulate the request_irq behaviour on top of s390 extints
+ */
+static void kvm_extint_handler(u16 code)
+{
+	void *data = (void *) *(long *) __LC_PFAULT_INTPARM;
+
+	vring_interrupt(0, data);
+}
+
+/*
+ * Init function for virtio
+ * devices are in a single page above top of "normal" mem
+ */
+static int __init kvm_devices_init(void)
+{
+	if (!MACHINE_IS_KVM)
+		return -ENODEV;
+
+	if (device_register(&kvm_root) != 0)
+		panic("Could not register kvm root");
+
+	if (add_shared_memory((max_pfn) << PAGE_SHIFT, PAGE_SIZE)) {
+		device_unregister(&kvm_root);
+		return -ENOMEM;
+	}
+
+	kvm_devices  = (void *) (max_pfn << PAGE_SHIFT);
+
+	register_external_interrupt(0x1237, kvm_extint_handler);
+	ctl_set_bit(0, 9);
+
+	scan_devices();
+	return 0;
+}
+
+/*
+ * We do this after core stuff, but before the drivers.
+ */
+postcore_initcall(kvm_devices_init);
Index: kvm/drivers/s390/kvm/kvm_virtio.h
==================================================================--- /dev/null
+++ kvm/drivers/s390/kvm/kvm_virtio.h
@@ -0,0 +1,47 @@
+/*
+ * kvm_virtio.h - definition for virtio for kvm on s390
+ *
+ * Copyright IBM Corp. 2008
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License (version 2 only)
+ * as published by the Free Software Foundation.
+ *
+ *    Author(s): Christian Borntraeger <borntraeger at de.ibm.com>
+ */
+
+#ifndef __KVM_S390_VIRTIO_H
+#define __KVM_S390_VIRTIO_H
+
+struct kvm_device_desc {
+	/* The device type: console, network, disk etc.  Type 0 terminates. */
+	__u8 type;
+	/* The number of virtqueues (first in config array) */
+	__u8 num_vq;
+	/*
+	 * The number of bytes of feature bits.  Multiply by 2: one for host
+	 * features and one for guest acknowledgements.
+	 */
+	__u8 feature_len;
+	/* The number of bytes of the config array after virtqueues. */
+	__u8 config_len;
+	/* A status byte, written by the Guest. */
+	__u8 status;
+	__u8 config[0];
+};
+
+/*
+ * This is how we expect the device configuration field for a virtqueue
+ * to be laid out in config space.
+ */
+struct kvm_vqconfig {
+	/* The token returned with an interrupt. Set by the guest */
+	__u64 token;
+	/* The address of the virtio ring */
+	__u64 address;
+	/* The number of entries in the virtio_ring */
+	__u16 num;
+
+};
+#endif
+
Index: kvm/include/asm-s390/kvm_para.h
==================================================================---
kvm.orig/include/asm-s390/kvm_para.h
+++ kvm/include/asm-s390/kvm_para.h
@@ -14,14 +14,134 @@
 #define __S390_KVM_PARA_H
 
 /*
- * No hypercalls for KVM on s390
+ * Hypercalls for KVM on s390. The calling convention is similar to the
+ * s390 ABI, so we use R2-R6 for parameters 1-5. In addition we use R1
+ * as hypercall number and R7 as parameter 6. The return value is
+ * written to R2. We use the diagnose instruction as hypercall. To avoid
+ * conflicts with existing diagnoses for LPAR and z/VM, we do not use
+ * the instruction encoded number, but specify the number in R1 and
+ * use 0x500 as KVM hypercall
+ *
+ * Copyright IBM Corp. 2007,2008
+ * Author(s): Christian Borntraeger <borntraeger at de.ibm.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.
  */
 
+static inline long kvm_hypercall0(unsigned long nr)
+{
+	register unsigned long __nr asm("1") = nr;
+	register long __rc asm("2");
+
+	asm volatile ("diag 2,4,0x500\n"
+		      : "=d" (__rc) : "d" (__nr): "memory",
"cc");
+	return __rc;
+}
+
+static inline long kvm_hypercall1(unsigned long nr, unsigned long p1)
+{
+	register unsigned long __nr asm("1") = nr;
+	register unsigned long __p1 asm("2") = p1;
+	register long __rc asm("2");
+
+	asm volatile ("diag 2,4,0x500\n"
+		      : "=d" (__rc) : "d" (__nr), "0" (__p1) :
"memory", "cc");
+	return __rc;
+}
+
+static inline long kvm_hypercall2(unsigned long nr, unsigned long p1,
+			       unsigned long p2)
+{
+	register unsigned long __nr asm("1") = nr;
+	register unsigned long __p1 asm("2") = p1;
+	register unsigned long __p2 asm("3") = p2;
+	register long __rc asm("2");
+
+	asm volatile ("diag 2,4,0x500\n"
+		      : "=d" (__rc) : "d" (__nr), "0" (__p1),
"d" (__p2)
+		      : "memory", "cc");
+	return __rc;
+}
+
+static inline long kvm_hypercall3(unsigned long nr, unsigned long p1,
+			       unsigned long p2, unsigned long p3)
+{
+	register unsigned long __nr asm("1") = nr;
+	register unsigned long __p1 asm("2") = p1;
+	register unsigned long __p2 asm("3") = p2;
+	register unsigned long __p3 asm("4") = p3;
+	register long __rc asm("2");
+
+	asm volatile ("diag 2,4,0x500\n"
+		      : "=d" (__rc) : "d" (__nr), "0" (__p1),
"d" (__p2),
+			"d" (__p3) : "memory", "cc");
+	return __rc;
+}
+
+
+static inline long kvm_hypercall4(unsigned long nr, unsigned long p1,
+			       unsigned long p2, unsigned long p3,
+			       unsigned long p4)
+{
+	register unsigned long __nr asm("1") = nr;
+	register unsigned long __p1 asm("2") = p1;
+	register unsigned long __p2 asm("3") = p2;
+	register unsigned long __p3 asm("4") = p3;
+	register unsigned long __p4 asm("5") = p4;
+	register long __rc asm("2");
+
+	asm volatile ("diag 2,4,0x500\n"
+		      : "=d" (__rc) : "d" (__nr), "0" (__p1),
"d" (__p2),
+			"d" (__p3), "d" (__p4) : "memory",
"cc");
+	return __rc;
+}
+
+static inline long kvm_hypercall5(unsigned long nr, unsigned long p1,
+			       unsigned long p2, unsigned long p3,
+			       unsigned long p4, unsigned long p5)
+{
+	register unsigned long __nr asm("1") = nr;
+	register unsigned long __p1 asm("2") = p1;
+	register unsigned long __p2 asm("3") = p2;
+	register unsigned long __p3 asm("4") = p3;
+	register unsigned long __p4 asm("5") = p4;
+	register unsigned long __p5 asm("6") = p5;
+	register long __rc asm("2");
+
+	asm volatile ("diag 2,4,0x500\n"
+		      : "=d" (__rc) : "d" (__nr), "0" (__p1),
"d" (__p2),
+			"d" (__p3), "d" (__p4), "d" (__p5)  :
"memory", "cc");
+	return __rc;
+}
+
+static inline long kvm_hypercall6(unsigned long nr, unsigned long p1,
+			       unsigned long p2, unsigned long p3,
+			       unsigned long p4, unsigned long p5,
+			       unsigned long p6)
+{
+	register unsigned long __nr asm("1") = nr;
+	register unsigned long __p1 asm("2") = p1;
+	register unsigned long __p2 asm("3") = p2;
+	register unsigned long __p3 asm("4") = p3;
+	register unsigned long __p4 asm("5") = p4;
+	register unsigned long __p5 asm("6") = p5;
+	register unsigned long __p6 asm("7") = p6;
+	register long __rc asm("2");
+
+	asm volatile ("diag 2,4,0x500\n"
+		      : "=d" (__rc) : "d" (__nr), "0" (__p1),
"d" (__p2),
+			"d" (__p3), "d" (__p4), "d" (__p5),
"d" (__p6)
+		      : "memory", "cc");
+	return __rc;
+}
+
+/* kvm on s390 is always paravirtualization enabled */
 static inline int kvm_para_available(void)
 {
-	return 0;
+	return 1;
 }
 
+/* No feature bits are currently assigned for kvm on s390 */
 static inline unsigned int kvm_arch_para_features(void)
 {
 	return 0;

On Friday 21 March 2008 03:25:28 Carsten Otte wrote:
> +static void kvm_set_status(struct virtio_device *vdev, u8 status)
> +{
> +	BUG_ON(!status);
> +	to_kvmdev(vdev)->desc->status = status;
> +}
> +
> +/*
> + * To reset the device, we (ab)use the NOTIFY hypercall, with the
descriptor
> + * address of the device.  The Host will zero the status and all the 
> + * features. 
> + */
> +static void kvm_reset(struct virtio_device *vdev)
> +{
> +	unsigned long offset = (void *)to_kvmdev(vdev)->desc - kvm_devices;
> +
> +	kvm_hypercall1(1237, (max_pfn<<PAGE_SHIFT) + offset);
> +}
I'd recommend a hypercall after set_status, as well as reset.  The
reason lguest doesn't do this is that we don't do feature negotiation
(assuming guest kernel matches host kernel).  In general, the host
needs to know when the VIRTIO_CONFIG_S_DRIVER_OK is set so it can see
what features the guest driver accepted.

Overloading the notify hypercall is kind of a hack too, but it works so
no real need to change that.
> + * The root device for the kvm virtio devices.
> + * This makes them appear as /sys/devices/kvm/0,1,2 not
/sys/devices/0,1,2.
> + */ 
> +static struct device kvm_root = {
> +	.parent = NULL,
> +	.bus_id = "kvm_s390",
> +};
You mean /sys/devices/kvm_s390/0,1,2?
> +static int __init kvm_devices_init(void)
> +{
> +	if (!MACHINE_IS_KVM)
> +		return -ENODEV;
> +
> +	if (device_register(&kvm_root) != 0)
> +		panic("Could not register kvm root");
> +
> +	if (add_shared_memory((max_pfn) << PAGE_SHIFT, PAGE_SIZE)) {
> +		device_unregister(&kvm_root);
> +		return -ENOMEM;
> +	}
Hmm, panic on device_register fail, but -ENOMEM on add_shared_memory fail?
My theory was that since this is boot time, panic() is the right thing.

Cheers,
Rusty.

Carsten Otte wrote:
> Currently we dont have PCI on s390. Making virtio_pci usable for s390 seems
> more complicated that providing an own stub. This virtio stub is similar to
> the lguest one, the memory for the descriptors and the device detection is
made
> via additional mapped memory on top of the guest storage. We use an
external
> interrupt with extint code 1237 for host->guest notification. 
>   
So, sanity won in the end.

-- 
Any sufficiently difficult bug is indistinguishable from a feature.