4 * This module allows virtio devices to be used over a virtual PCI device.
5 * This can be used with QEMU based VMMs like KVM or Xen.
7 * Copyright IBM Corp. 2007
10 * Anthony Liguori <aliguori@us.ibm.com>
12 * This work is licensed under the terms of the GNU GPL, version 2 or later.
13 * See the COPYING file in the top-level directory.
17 #include <linux/module.h>
18 #include <linux/list.h>
19 #include <linux/pci.h>
20 #include <linux/interrupt.h>
21 #include <linux/virtio.h>
22 #include <linux/virtio_config.h>
23 #include <linux/virtio_ring.h>
24 #include <linux/virtio_pci.h>
25 #include <linux/highmem.h>
26 #include <linux/spinlock.h>
28 MODULE_AUTHOR("Anthony Liguori <aliguori@us.ibm.com>");
29 MODULE_DESCRIPTION("virtio-pci");
30 MODULE_LICENSE("GPL");
33 /* Our device structure */
34 struct virtio_pci_device
36 struct virtio_device vdev;
37 struct pci_dev *pci_dev;
39 /* the IO mapping for the PCI config space */
42 /* a list of queues so we can dispatch IRQs */
44 struct list_head virtqueues;
47 struct virtio_pci_vq_info
49 /* the actual virtqueue */
52 /* the number of entries in the queue */
55 /* the index of the queue */
58 /* the virtual address of the ring queue */
61 /* the list node for the virtqueues list */
62 struct list_head node;
65 /* Qumranet donated their vendor ID for devices 0x1000 thru 0x10FF. */
66 static struct pci_device_id virtio_pci_id_table[] = {
67 { 0x1af4, PCI_ANY_ID, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0 },
71 MODULE_DEVICE_TABLE(pci, virtio_pci_id_table);
73 /* A PCI device has it's own struct device and so does a virtio device so
74 * we create a place for the virtio devices to show up in sysfs. I think it
75 * would make more sense for virtio to not insist on having it's own device. */
76 static struct device virtio_pci_root = {
78 .bus_id = "virtio-pci",
81 /* Convert a generic virtio device to our structure */
82 static struct virtio_pci_device *to_vp_device(struct virtio_device *vdev)
84 return container_of(vdev, struct virtio_pci_device, vdev);
87 /* virtio config->get_features() implementation */
88 static u32 vp_get_features(struct virtio_device *vdev)
90 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
92 /* When someone needs more than 32 feature bits, we'll need to
93 * steal a bit to indicate that the rest are somewhere else. */
94 return ioread32(vp_dev->ioaddr + VIRTIO_PCI_HOST_FEATURES);
97 /* virtio config->set_features() implementation */
98 static void vp_set_features(struct virtio_device *vdev, u32 features)
100 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
102 iowrite32(features, vp_dev->ioaddr + VIRTIO_PCI_GUEST_FEATURES);
105 /* virtio config->get() implementation */
106 static void vp_get(struct virtio_device *vdev, unsigned offset,
107 void *buf, unsigned len)
109 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
110 void __iomem *ioaddr = vp_dev->ioaddr + VIRTIO_PCI_CONFIG + offset;
114 for (i = 0; i < len; i++)
115 ptr[i] = ioread8(ioaddr + i);
118 /* the config->set() implementation. it's symmetric to the config->get()
120 static void vp_set(struct virtio_device *vdev, unsigned offset,
121 const void *buf, unsigned len)
123 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
124 void __iomem *ioaddr = vp_dev->ioaddr + VIRTIO_PCI_CONFIG + offset;
128 for (i = 0; i < len; i++)
129 iowrite8(ptr[i], ioaddr + i);
132 /* config->{get,set}_status() implementations */
133 static u8 vp_get_status(struct virtio_device *vdev)
135 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
136 return ioread8(vp_dev->ioaddr + VIRTIO_PCI_STATUS);
139 static void vp_set_status(struct virtio_device *vdev, u8 status)
141 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
142 /* We should never be setting status to 0. */
144 iowrite8(status, vp_dev->ioaddr + VIRTIO_PCI_STATUS);
147 static void vp_reset(struct virtio_device *vdev)
149 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
150 /* 0 status means a reset. */
151 iowrite8(0, vp_dev->ioaddr + VIRTIO_PCI_STATUS);
154 /* the notify function used when creating a virt queue */
155 static void vp_notify(struct virtqueue *vq)
157 struct virtio_pci_device *vp_dev = to_vp_device(vq->vdev);
158 struct virtio_pci_vq_info *info = vq->priv;
160 /* we write the queue's selector into the notification register to
161 * signal the other end */
162 iowrite16(info->queue_index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_NOTIFY);
165 /* A small wrapper to also acknowledge the interrupt when it's handled.
166 * I really need an EIO hook for the vring so I can ack the interrupt once we
167 * know that we'll be handling the IRQ but before we invoke the callback since
168 * the callback may notify the host which results in the host attempting to
169 * raise an interrupt that we would then mask once we acknowledged the
171 static irqreturn_t vp_interrupt(int irq, void *opaque)
173 struct virtio_pci_device *vp_dev = opaque;
174 struct virtio_pci_vq_info *info;
175 irqreturn_t ret = IRQ_NONE;
179 /* reading the ISR has the effect of also clearing it so it's very
180 * important to save off the value. */
181 isr = ioread8(vp_dev->ioaddr + VIRTIO_PCI_ISR);
183 /* It's definitely not us if the ISR was not high */
187 /* Configuration change? Tell driver if it wants to know. */
188 if (isr & VIRTIO_PCI_ISR_CONFIG) {
189 struct virtio_driver *drv;
190 drv = container_of(vp_dev->vdev.dev.driver,
191 struct virtio_driver, driver);
193 if (drv->config_changed)
194 drv->config_changed(&vp_dev->vdev);
197 spin_lock_irqsave(&vp_dev->lock, flags);
198 list_for_each_entry(info, &vp_dev->virtqueues, node) {
199 if (vring_interrupt(irq, info->vq) == IRQ_HANDLED)
202 spin_unlock_irqrestore(&vp_dev->lock, flags);
207 /* the config->find_vq() implementation */
208 static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
209 void (*callback)(struct virtqueue *vq))
211 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
212 struct virtio_pci_vq_info *info;
213 struct virtqueue *vq;
218 /* Select the queue we're interested in */
219 iowrite16(index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_SEL);
221 /* Check if queue is either not available or already active. */
222 num = ioread16(vp_dev->ioaddr + VIRTIO_PCI_QUEUE_NUM);
223 if (!num || ioread32(vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN))
224 return ERR_PTR(-ENOENT);
226 /* allocate and fill out our structure the represents an active
228 info = kmalloc(sizeof(struct virtio_pci_vq_info), GFP_KERNEL);
230 return ERR_PTR(-ENOMEM);
232 info->queue_index = index;
235 info->queue = kzalloc(PAGE_ALIGN(vring_size(num,PAGE_SIZE)), GFP_KERNEL);
236 if (info->queue == NULL) {
241 /* activate the queue */
242 iowrite32(virt_to_phys(info->queue) >> PAGE_SHIFT,
243 vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
245 /* create the vring */
246 vq = vring_new_virtqueue(info->num, vdev, info->queue,
247 vp_notify, callback);
250 goto out_activate_queue;
256 spin_lock_irqsave(&vp_dev->lock, flags);
257 list_add(&info->node, &vp_dev->virtqueues);
258 spin_unlock_irqrestore(&vp_dev->lock, flags);
263 iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
270 /* the config->del_vq() implementation */
271 static void vp_del_vq(struct virtqueue *vq)
273 struct virtio_pci_device *vp_dev = to_vp_device(vq->vdev);
274 struct virtio_pci_vq_info *info = vq->priv;
277 spin_lock_irqsave(&vp_dev->lock, flags);
278 list_del(&info->node);
279 spin_unlock_irqrestore(&vp_dev->lock, flags);
281 vring_del_virtqueue(vq);
283 /* Select and deactivate the queue */
284 iowrite16(info->queue_index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_SEL);
285 iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
291 static struct virtio_config_ops virtio_pci_config_ops = {
294 .get_status = vp_get_status,
295 .set_status = vp_set_status,
297 .find_vq = vp_find_vq,
299 .get_features = vp_get_features,
300 .set_features = vp_set_features,
303 /* the PCI probing function */
304 static int __devinit virtio_pci_probe(struct pci_dev *pci_dev,
305 const struct pci_device_id *id)
307 struct virtio_pci_device *vp_dev;
310 /* We only own devices >= 0x1000 and <= 0x103f: leave the rest. */
311 if (pci_dev->device < 0x1000 || pci_dev->device > 0x103f)
314 if (pci_dev->revision != VIRTIO_PCI_ABI_VERSION) {
315 printk(KERN_ERR "virtio_pci: expected ABI version %d, got %d\n",
316 VIRTIO_PCI_ABI_VERSION, pci_dev->revision);
320 /* allocate our structure and fill it out */
321 vp_dev = kzalloc(sizeof(struct virtio_pci_device), GFP_KERNEL);
325 vp_dev->vdev.dev.parent = &virtio_pci_root;
326 vp_dev->vdev.config = &virtio_pci_config_ops;
327 vp_dev->pci_dev = pci_dev;
328 INIT_LIST_HEAD(&vp_dev->virtqueues);
329 spin_lock_init(&vp_dev->lock);
331 /* enable the device */
332 err = pci_enable_device(pci_dev);
336 err = pci_request_regions(pci_dev, "virtio-pci");
338 goto out_enable_device;
340 vp_dev->ioaddr = pci_iomap(pci_dev, 0, 0);
341 if (vp_dev->ioaddr == NULL)
342 goto out_req_regions;
344 pci_set_drvdata(pci_dev, vp_dev);
346 /* we use the subsystem vendor/device id as the virtio vendor/device
347 * id. this allows us to use the same PCI vendor/device id for all
348 * virtio devices and to identify the particular virtio driver by
349 * the subsytem ids */
350 vp_dev->vdev.id.vendor = pci_dev->subsystem_vendor;
351 vp_dev->vdev.id.device = pci_dev->subsystem_device;
353 /* register a handler for the queue with the PCI device's interrupt */
354 err = request_irq(vp_dev->pci_dev->irq, vp_interrupt, IRQF_SHARED,
355 vp_dev->vdev.dev.bus_id, vp_dev);
357 goto out_set_drvdata;
359 /* finally register the virtio device */
360 err = register_virtio_device(&vp_dev->vdev);
367 free_irq(pci_dev->irq, vp_dev);
369 pci_set_drvdata(pci_dev, NULL);
370 pci_iounmap(pci_dev, vp_dev->ioaddr);
372 pci_release_regions(pci_dev);
374 pci_disable_device(pci_dev);
380 static void __devexit virtio_pci_remove(struct pci_dev *pci_dev)
382 struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
384 unregister_virtio_device(&vp_dev->vdev);
385 free_irq(pci_dev->irq, vp_dev);
386 pci_set_drvdata(pci_dev, NULL);
387 pci_iounmap(pci_dev, vp_dev->ioaddr);
388 pci_release_regions(pci_dev);
389 pci_disable_device(pci_dev);
394 static int virtio_pci_suspend(struct pci_dev *pci_dev, pm_message_t state)
396 pci_save_state(pci_dev);
397 pci_set_power_state(pci_dev, PCI_D3hot);
401 static int virtio_pci_resume(struct pci_dev *pci_dev)
403 pci_restore_state(pci_dev);
404 pci_set_power_state(pci_dev, PCI_D0);
409 static struct pci_driver virtio_pci_driver = {
410 .name = "virtio-pci",
411 .id_table = virtio_pci_id_table,
412 .probe = virtio_pci_probe,
413 .remove = virtio_pci_remove,
415 .suspend = virtio_pci_suspend,
416 .resume = virtio_pci_resume,
420 static int __init virtio_pci_init(void)
424 err = device_register(&virtio_pci_root);
428 err = pci_register_driver(&virtio_pci_driver);
430 device_unregister(&virtio_pci_root);
435 module_init(virtio_pci_init);
437 static void __exit virtio_pci_exit(void)
439 device_unregister(&virtio_pci_root);
440 pci_unregister_driver(&virtio_pci_driver);
443 module_exit(virtio_pci_exit);