[mirror_ubuntu-artful-kernel.git] / drivers / virtio / virtio_pci_common.c

/*
 * Virtio PCI driver - common functionality for all device versions
 *
 * This module allows virtio devices to be used over a virtual PCI device.
 * This can be used with QEMU based VMMs like KVM or Xen.
 *
 * Copyright IBM Corp. 2007
 * Copyright Red Hat, Inc. 2014
 *
 * Authors:
 *  Anthony Liguori  <aliguori@us.ibm.com>
 *  Rusty Russell <rusty@rustcorp.com.au>
 *  Michael S. Tsirkin <mst@redhat.com>
 *
 * This work is licensed under the terms of the GNU GPL, version 2 or later.
 * See the COPYING file in the top-level directory.
 *
 */

#include "virtio_pci_common.h"

static bool force_legacy = false;

#if IS_ENABLED(CONFIG_VIRTIO_PCI_LEGACY)
module_param(force_legacy, bool, 0444);
MODULE_PARM_DESC(force_legacy,
		 "Force legacy mode for transitional virtio 1 devices");
#endif

/* wait for pending irq handlers */
void vp_synchronize_vectors(struct virtio_device *vdev)
{
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	int i;

	synchronize_irq(pci_irq_vector(vp_dev->pci_dev, 0));
	for (i = 1; i < vp_dev->msix_vectors; i++)
		synchronize_irq(pci_irq_vector(vp_dev->pci_dev, i));
}

/* the notify function used when creating a virt queue */
bool vp_notify(struct virtqueue *vq)
{
	/* we write the queue's selector into the notification register to
	 * signal the other end */
	iowrite16(vq->index, (void __iomem *)vq->priv);
	return true;
}

/* Handle a configuration change: Tell driver if it wants to know. */
static irqreturn_t vp_config_changed(int irq, void *opaque)
{
	struct virtio_pci_device *vp_dev = opaque;

	virtio_config_changed(&vp_dev->vdev);
	return IRQ_HANDLED;
}

/* Notify all virtqueues on an interrupt. */
static irqreturn_t vp_vring_interrupt(int irq, void *opaque)
{
	struct virtio_pci_device *vp_dev = opaque;
	irqreturn_t ret = IRQ_NONE;
	struct virtqueue *vq;

	list_for_each_entry(vq, &vp_dev->vdev.vqs, list) {
		if (vq->callback && vring_interrupt(irq, vq) == IRQ_HANDLED)
			ret = IRQ_HANDLED;
	}

	return ret;
}

/* A small wrapper to also acknowledge the interrupt when it's handled.
 * I really need an EIO hook for the vring so I can ack the interrupt once we
 * know that we'll be handling the IRQ but before we invoke the callback since
 * the callback may notify the host which results in the host attempting to
 * raise an interrupt that we would then mask once we acknowledged the
 * interrupt. */
static irqreturn_t vp_interrupt(int irq, void *opaque)
{
	struct virtio_pci_device *vp_dev = opaque;
	u8 isr;

	/* reading the ISR has the effect of also clearing it so it's very
	 * important to save off the value. */
	isr = ioread8(vp_dev->isr);

	/* It's definitely not us if the ISR was not high */
	if (!isr)
		return IRQ_NONE;

	/* Configuration change?  Tell driver if it wants to know. */
	if (isr & VIRTIO_PCI_ISR_CONFIG)
		vp_config_changed(irq, opaque);

	return vp_vring_interrupt(irq, opaque);
}

static void vp_remove_vqs(struct virtio_device *vdev)
{
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	struct virtqueue *vq, *n;

	list_for_each_entry_safe(vq, n, &vdev->vqs, list) {
		if (vp_dev->msix_vector_map) {
			int v = vp_dev->msix_vector_map[vq->index];

			if (v != VIRTIO_MSI_NO_VECTOR)
				free_irq(pci_irq_vector(vp_dev->pci_dev, v),
					vq);
		}
		vp_dev->del_vq(vq);
	}
}

/* the config->del_vqs() implementation */
void vp_del_vqs(struct virtio_device *vdev)
{
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	int i;

	if (WARN_ON_ONCE(list_empty_careful(&vdev->vqs)))
		return;

	vp_remove_vqs(vdev);

	if (vp_dev->pci_dev->msix_enabled) {
		for (i = 0; i < vp_dev->msix_vectors; i++)
			free_cpumask_var(vp_dev->msix_affinity_masks[i]);

		/* Disable the vector used for configuration */
		vp_dev->config_vector(vp_dev, VIRTIO_MSI_NO_VECTOR);

		kfree(vp_dev->msix_affinity_masks);
		kfree(vp_dev->msix_names);
		kfree(vp_dev->msix_vector_map);
	}

	free_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_dev);
	pci_free_irq_vectors(vp_dev->pci_dev);
}

static int vp_find_vqs_msix(struct virtio_device *vdev, unsigned nvqs,
		struct virtqueue *vqs[], vq_callback_t *callbacks[],
		const char * const names[], struct irq_affinity *desc)
{
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	const char *name = dev_name(&vp_dev->vdev.dev);
	int i, err = -ENOMEM, allocated_vectors, nvectors;
	unsigned flags = PCI_IRQ_MSIX;
	bool shared = false;
	u16 msix_vec;

	if (desc) {
		flags |= PCI_IRQ_AFFINITY;
		desc->pre_vectors++; /* virtio config vector */
	}

	nvectors = 1;
	for (i = 0; i < nvqs; i++)
		if (callbacks[i])
			nvectors++;

	/* Try one vector per queue first. */
	err = pci_alloc_irq_vectors_affinity(vp_dev->pci_dev, nvectors,
			nvectors, flags, desc);
	if (err < 0) {
		/* Fallback to one vector for config, one shared for queues. */
		shared = true;
		err = pci_alloc_irq_vectors(vp_dev->pci_dev, 2, 2,
				PCI_IRQ_MSIX);
		if (err < 0)
			return err;
	}
	if (err < 0)
		return err;

	vp_dev->msix_vectors = nvectors;
	vp_dev->msix_names = kmalloc_array(nvectors,
			sizeof(*vp_dev->msix_names), GFP_KERNEL);
	if (!vp_dev->msix_names)
		goto out_free_irq_vectors;

	vp_dev->msix_affinity_masks = kcalloc(nvectors,
			sizeof(*vp_dev->msix_affinity_masks), GFP_KERNEL);
	if (!vp_dev->msix_affinity_masks)
		goto out_free_msix_names;

	for (i = 0; i < nvectors; ++i) {
		if (!alloc_cpumask_var(&vp_dev->msix_affinity_masks[i],
				GFP_KERNEL))
			goto out_free_msix_affinity_masks;
	}

	/* Set the vector used for configuration */
	snprintf(vp_dev->msix_names[0], sizeof(*vp_dev->msix_names),
		 "%s-config", name);
	err = request_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_config_changed,
			0, vp_dev->msix_names[0], vp_dev);
	if (err)
		goto out_free_msix_affinity_masks;

	/* Verify we had enough resources to assign the vector */
	if (vp_dev->config_vector(vp_dev, 0) == VIRTIO_MSI_NO_VECTOR) {
		err = -EBUSY;
		goto out_free_config_irq;
	}

	vp_dev->msix_vector_map = kmalloc_array(nvqs,
			sizeof(*vp_dev->msix_vector_map), GFP_KERNEL);
	if (!vp_dev->msix_vector_map)
		goto out_disable_config_irq;

	allocated_vectors = 1; /* vector 0 is the config interrupt */
	for (i = 0; i < nvqs; ++i) {
		if (!names[i]) {
			vqs[i] = NULL;
			continue;
		}

		if (callbacks[i])
			msix_vec = allocated_vectors;
		else
			msix_vec = VIRTIO_MSI_NO_VECTOR;

		vqs[i] = vp_dev->setup_vq(vp_dev, i, callbacks[i], names[i],
				msix_vec);
		if (IS_ERR(vqs[i])) {
			err = PTR_ERR(vqs[i]);
			goto out_remove_vqs;
		}

		if (msix_vec == VIRTIO_MSI_NO_VECTOR) {
			vp_dev->msix_vector_map[i] = VIRTIO_MSI_NO_VECTOR;
			continue;
		}

		snprintf(vp_dev->msix_names[i + 1],
			 sizeof(*vp_dev->msix_names), "%s-%s",
			 dev_name(&vp_dev->vdev.dev), names[i]);
		err = request_irq(pci_irq_vector(vp_dev->pci_dev, msix_vec),
				  vring_interrupt, IRQF_SHARED,
				  vp_dev->msix_names[i + 1], vqs[i]);
		if (err) {
			/* don't free this irq on error */
			vp_dev->msix_vector_map[i] = VIRTIO_MSI_NO_VECTOR;
			goto out_remove_vqs;
		}
		vp_dev->msix_vector_map[i] = msix_vec;

		/*
		 * Use a different vector for each queue if they are available,
		 * else share the same vector for all VQs.
		 */
		if (!shared)
			allocated_vectors++;
	}

	return 0;

out_remove_vqs:
	vp_remove_vqs(vdev);
	kfree(vp_dev->msix_vector_map);
out_disable_config_irq:
	vp_dev->config_vector(vp_dev, VIRTIO_MSI_NO_VECTOR);
out_free_config_irq:
	free_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_dev);
out_free_msix_affinity_masks:
	for (i = 0; i < nvectors; i++) {
		if (vp_dev->msix_affinity_masks[i])
			free_cpumask_var(vp_dev->msix_affinity_masks[i]);
	}
	kfree(vp_dev->msix_affinity_masks);
out_free_msix_names:
	kfree(vp_dev->msix_names);
out_free_irq_vectors:
	pci_free_irq_vectors(vp_dev->pci_dev);
	return err;
}

static int vp_find_vqs_intx(struct virtio_device *vdev, unsigned nvqs,
		struct virtqueue *vqs[], vq_callback_t *callbacks[],
		const char * const names[])
{
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	int i, err;

	err = request_irq(vp_dev->pci_dev->irq, vp_interrupt, IRQF_SHARED,
			dev_name(&vdev->dev), vp_dev);
	if (err)
		return err;

	for (i = 0; i < nvqs; ++i) {
		if (!names[i]) {
			vqs[i] = NULL;
			continue;
		}
		vqs[i] = vp_dev->setup_vq(vp_dev, i, callbacks[i], names[i],
				VIRTIO_MSI_NO_VECTOR);
		if (IS_ERR(vqs[i])) {
			err = PTR_ERR(vqs[i]);
			goto out_remove_vqs;
		}
	}

	return 0;

out_remove_vqs:
	vp_remove_vqs(vdev);
	free_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_dev);
	return err;
}

/* the config->find_vqs() implementation */
int vp_find_vqs(struct virtio_device *vdev, unsigned nvqs,
		struct virtqueue *vqs[], vq_callback_t *callbacks[],
		const char * const names[], struct irq_affinity *desc)
{
	int err;

	err = vp_find_vqs_msix(vdev, nvqs, vqs, callbacks, names, desc);
	if (!err)
		return 0;
	return vp_find_vqs_intx(vdev, nvqs, vqs, callbacks, names);
}

const char *vp_bus_name(struct virtio_device *vdev)
{
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);

	return pci_name(vp_dev->pci_dev);
}

/* Setup the affinity for a virtqueue:
 * - force the affinity for per vq vector
 * - OR over all affinities for shared MSI
 * - ignore the affinity request if we're using INTX
 */
int vp_set_vq_affinity(struct virtqueue *vq, int cpu)
{
	struct virtio_device *vdev = vq->vdev;
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);

	if (!vq->callback)
		return -EINVAL;

	if (vp_dev->pci_dev->msix_enabled) {
		int vec = vp_dev->msix_vector_map[vq->index];
		struct cpumask *mask = vp_dev->msix_affinity_masks[vec];
		unsigned int irq = pci_irq_vector(vp_dev->pci_dev, vec);

		if (cpu == -1)
			irq_set_affinity_hint(irq, NULL);
		else {
			cpumask_clear(mask);
			cpumask_set_cpu(cpu, mask);
			irq_set_affinity_hint(irq, mask);
		}
	}
	return 0;
}

const struct cpumask *vp_get_vq_affinity(struct virtio_device *vdev, int index)
{
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	unsigned int *map = vp_dev->msix_vector_map;

	if (!map || map[index] == VIRTIO_MSI_NO_VECTOR)
		return NULL;

	return pci_irq_get_affinity(vp_dev->pci_dev, map[index]);
}

#ifdef CONFIG_PM_SLEEP
static int virtio_pci_freeze(struct device *dev)
{
	struct pci_dev *pci_dev = to_pci_dev(dev);
	struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
	int ret;

	ret = virtio_device_freeze(&vp_dev->vdev);

	if (!ret)
		pci_disable_device(pci_dev);
	return ret;
}

static int virtio_pci_restore(struct device *dev)
{
	struct pci_dev *pci_dev = to_pci_dev(dev);
	struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
	int ret;

	ret = pci_enable_device(pci_dev);
	if (ret)
		return ret;

	pci_set_master(pci_dev);
	return virtio_device_restore(&vp_dev->vdev);
}

static const struct dev_pm_ops virtio_pci_pm_ops = {
	SET_SYSTEM_SLEEP_PM_OPS(virtio_pci_freeze, virtio_pci_restore)
};
#endif


/* Qumranet donated their vendor ID for devices 0x1000 thru 0x10FF. */
static const struct pci_device_id virtio_pci_id_table[] = {
	{ PCI_DEVICE(PCI_VENDOR_ID_REDHAT_QUMRANET, PCI_ANY_ID) },
	{ 0 }
};

MODULE_DEVICE_TABLE(pci, virtio_pci_id_table);

static void virtio_pci_release_dev(struct device *_d)
{
	struct virtio_device *vdev = dev_to_virtio(_d);
	struct virtio_pci_device *vp_dev = to_vp_device(vdev);

	/* As struct device is a kobject, it's not safe to
	 * free the memory (including the reference counter itself)
	 * until it's release callback. */
	kfree(vp_dev);
}

static int virtio_pci_probe(struct pci_dev *pci_dev,
			    const struct pci_device_id *id)
{
	struct virtio_pci_device *vp_dev;
	int rc;

	/* allocate our structure and fill it out */
	vp_dev = kzalloc(sizeof(struct virtio_pci_device), GFP_KERNEL);
	if (!vp_dev)
		return -ENOMEM;

	pci_set_drvdata(pci_dev, vp_dev);
	vp_dev->vdev.dev.parent = &pci_dev->dev;
	vp_dev->vdev.dev.release = virtio_pci_release_dev;
	vp_dev->pci_dev = pci_dev;

	/* enable the device */
	rc = pci_enable_device(pci_dev);
	if (rc)
		goto err_enable_device;

	if (force_legacy) {
		rc = virtio_pci_legacy_probe(vp_dev);
		/* Also try modern mode if we can't map BAR0 (no IO space). */
		if (rc == -ENODEV || rc == -ENOMEM)
			rc = virtio_pci_modern_probe(vp_dev);
		if (rc)
			goto err_probe;
	} else {
		rc = virtio_pci_modern_probe(vp_dev);
		if (rc == -ENODEV)
			rc = virtio_pci_legacy_probe(vp_dev);
		if (rc)
			goto err_probe;
	}

	pci_set_master(pci_dev);

	rc = register_virtio_device(&vp_dev->vdev);
	if (rc)
		goto err_register;

	return 0;

err_register:
	if (vp_dev->ioaddr)
	     virtio_pci_legacy_remove(vp_dev);
	else
	     virtio_pci_modern_remove(vp_dev);
err_probe:
	pci_disable_device(pci_dev);
err_enable_device:
	kfree(vp_dev);
	return rc;
}

static void virtio_pci_remove(struct pci_dev *pci_dev)
{
	struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
	struct device *dev = get_device(&vp_dev->vdev.dev);

	unregister_virtio_device(&vp_dev->vdev);

	if (vp_dev->ioaddr)
		virtio_pci_legacy_remove(vp_dev);
	else
		virtio_pci_modern_remove(vp_dev);

	pci_disable_device(pci_dev);
	put_device(dev);
}

static struct pci_driver virtio_pci_driver = {
	.name		= "virtio-pci",
	.id_table	= virtio_pci_id_table,
	.probe		= virtio_pci_probe,
	.remove		= virtio_pci_remove,
#ifdef CONFIG_PM_SLEEP
	.driver.pm	= &virtio_pci_pm_ops,
#endif
};

module_pci_driver(virtio_pci_driver);

MODULE_AUTHOR("Anthony Liguori <aliguori@us.ibm.com>");
MODULE_DESCRIPTION("virtio-pci");
MODULE_LICENSE("GPL");
MODULE_VERSION("1");
Commit	Line	Data
3343660d	1	/*
a90fdce9	2	* Virtio PCI driver - common functionality for all device versions
3343660d AL	3	*
	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.
	6	*
	7	* Copyright IBM Corp. 2007
a90fdce9	8	* Copyright Red Hat, Inc. 2014
3343660d AL	9	*
	10	* Authors:
	11	* Anthony Liguori <aliguori@us.ibm.com>
a90fdce9 MT	12	* Rusty Russell <rusty@rustcorp.com.au>
a90fdce9 MT	13	* Michael S. Tsirkin <mst@redhat.com>
3343660d AL	14	*
	15	* This work is licensed under the terms of the GNU GPL, version 2 or later.
	16	* See the COPYING file in the top-level directory.
	17	*
	18	*/
	19
5f4c9760	20	#include "virtio_pci_common.h"
3343660d	21
ac399d8f MT	22	static bool force_legacy = false;
	23
	24	#if IS_ENABLED(CONFIG_VIRTIO_PCI_LEGACY)
	25	module_param(force_legacy, bool, 0444);
	26	MODULE_PARM_DESC(force_legacy,
	27	"Force legacy mode for transitional virtio 1 devices");
	28	#endif
	29
e6af578c	30	/* wait for pending irq handlers */
38eb4a29	31	void vp_synchronize_vectors(struct virtio_device *vdev)
e6af578c MT	32	{
	33	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	34	int i;
	35
07ec5148 CH	36	synchronize_irq(pci_irq_vector(vp_dev->pci_dev, 0));
07ec5148 CH	37	for (i = 1; i < vp_dev->msix_vectors; i++)
fa3a3279	38	synchronize_irq(pci_irq_vector(vp_dev->pci_dev, i));
e6af578c MT	39	}
e6af578c MT	40
3343660d	41	/* the notify function used when creating a virt queue */
38eb4a29	42	bool vp_notify(struct virtqueue *vq)
3343660d	43	{
3343660d AL	44	/* we write the queue's selector into the notification register to
3343660d AL	45	* signal the other end */
f30eaf4a	46	iowrite16(vq->index, (void __iomem *)vq->priv);
46f9c2b9	47	return true;
3343660d AL	48	}
3343660d AL	49
77cf5246 MT	50	/* Handle a configuration change: Tell driver if it wants to know. */
	51	static irqreturn_t vp_config_changed(int irq, void *opaque)
	52	{
	53	struct virtio_pci_device *vp_dev = opaque;
77cf5246	54
016c98c6	55	virtio_config_changed(&vp_dev->vdev);
77cf5246 MT	56	return IRQ_HANDLED;
	57	}
	58
	59	/* Notify all virtqueues on an interrupt. */
	60	static irqreturn_t vp_vring_interrupt(int irq, void *opaque)
	61	{
	62	struct virtio_pci_device *vp_dev = opaque;
77cf5246	63	irqreturn_t ret = IRQ_NONE;
5c34d002	64	struct virtqueue *vq;
77cf5246	65
5c34d002 CH	66	list_for_each_entry(vq, &vp_dev->vdev.vqs, list) {
5c34d002 CH	67	if (vq->callback && vring_interrupt(irq, vq) == IRQ_HANDLED)
77cf5246 MT	68	ret = IRQ_HANDLED;
77cf5246 MT	69	}
77cf5246 MT	70
	71	return ret;
	72	}
	73
3343660d AL	74	/* A small wrapper to also acknowledge the interrupt when it's handled.
	75	* I really need an EIO hook for the vring so I can ack the interrupt once we
	76	* know that we'll be handling the IRQ but before we invoke the callback since
	77	* the callback may notify the host which results in the host attempting to
	78	* raise an interrupt that we would then mask once we acknowledged the
	79	* interrupt. */
	80	static irqreturn_t vp_interrupt(int irq, void *opaque)
	81	{
	82	struct virtio_pci_device *vp_dev = opaque;
3343660d AL	83	u8 isr;
	84
	85	/* reading the ISR has the effect of also clearing it so it's very
	86	* important to save off the value. */
af535722	87	isr = ioread8(vp_dev->isr);
3343660d AL	88
	89	/* It's definitely not us if the ISR was not high */
	90	if (!isr)
	91	return IRQ_NONE;
	92
	93	/* Configuration change? Tell driver if it wants to know. */
77cf5246 MT	94	if (isr & VIRTIO_PCI_ISR_CONFIG)
77cf5246 MT	95	vp_config_changed(irq, opaque);
3343660d	96
77cf5246	97	return vp_vring_interrupt(irq, opaque);
3343660d AL	98	}
3343660d AL	99
07ec5148	100	static void vp_remove_vqs(struct virtio_device *vdev)
d2a7ddda	101	{
e969fed5	102	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
d2a7ddda MT	103	struct virtqueue vq, n;
d2a7ddda MT	104
e969fed5	105	list_for_each_entry_safe(vq, n, &vdev->vqs, list) {
5c34d002 CH	106	if (vp_dev->msix_vector_map) {
5c34d002 CH	107	int v = vp_dev->msix_vector_map[vq->index];
fa3a3279 CH	108
	109	if (v != VIRTIO_MSI_NO_VECTOR)
	110	free_irq(pci_irq_vector(vp_dev->pci_dev, v),
	111	vq);
	112	}
5c34d002	113	vp_dev->del_vq(vq);
e969fed5	114	}
07ec5148	115	}
82af8ce8	116
07ec5148 CH	117	/* the config->del_vqs() implementation */
	118	void vp_del_vqs(struct virtio_device *vdev)
	119	{
	120	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	121	int i;
66f2f555	122
07ec5148 CH	123	if (WARN_ON_ONCE(list_empty_careful(&vdev->vqs)))
07ec5148 CH	124	return;
66f2f555	125
07ec5148	126	vp_remove_vqs(vdev);
66f2f555	127
53a020c6	128	if (vp_dev->pci_dev->msix_enabled) {
07ec5148 CH	129	for (i = 0; i < vp_dev->msix_vectors; i++)
	130	free_cpumask_var(vp_dev->msix_affinity_masks[i]);
	131
66f2f555 CH	132	/* Disable the vector used for configuration */
	133	vp_dev->config_vector(vp_dev, VIRTIO_MSI_NO_VECTOR);
	134
07ec5148 CH	135	kfree(vp_dev->msix_affinity_masks);
	136	kfree(vp_dev->msix_names);
	137	kfree(vp_dev->msix_vector_map);
66f2f555 CH	138	}
66f2f555 CH	139
07ec5148 CH	140	free_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_dev);
07ec5148 CH	141	pci_free_irq_vectors(vp_dev->pci_dev);
d2a7ddda MT	142	}
d2a7ddda MT	143
a3cbec69	144	static int vp_find_vqs_msix(struct virtio_device *vdev, unsigned nvqs,
52a61516	145	struct virtqueue vqs[], vq_callback_t callbacks[],
fb5e31d9	146	const char * const names[], struct irq_affinity *desc)
d2a7ddda	147	{
e969fed5	148	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
07ec5148 CH	149	const char *name = dev_name(&vp_dev->vdev.dev);
07ec5148 CH	150	int i, err = -ENOMEM, allocated_vectors, nvectors;
fb5e31d9	151	unsigned flags = PCI_IRQ_MSIX;
52a61516	152	bool shared = false;
f68d2408	153	u16 msix_vec;
07ec5148	154
fb5e31d9 CH	155	if (desc) {
	156	flags \|= PCI_IRQ_AFFINITY;
	157	desc->pre_vectors++; /* virtio config vector */
	158	}
	159
07ec5148 CH	160	nvectors = 1;
	161	for (i = 0; i < nvqs; i++)
	162	if (callbacks[i])
	163	nvectors++;
82af8ce8	164
52a61516	165	/* Try one vector per queue first. */
fb5e31d9 CH	166	err = pci_alloc_irq_vectors_affinity(vp_dev->pci_dev, nvectors,
fb5e31d9 CH	167	nvectors, flags, desc);
52a61516 CH	168	if (err < 0) {
	169	/* Fallback to one vector for config, one shared for queues. */
	170	shared = true;
07ec5148 CH	171	err = pci_alloc_irq_vectors(vp_dev->pci_dev, 2, 2,
07ec5148 CH	172	PCI_IRQ_MSIX);
52a61516 CH	173	if (err < 0)
52a61516 CH	174	return err;
f68d2408	175	}
07ec5148 CH	176	if (err < 0)
07ec5148 CH	177	return err;
d2a7ddda	178
07ec5148 CH	179	vp_dev->msix_vectors = nvectors;
	180	vp_dev->msix_names = kmalloc_array(nvectors,
	181	sizeof(*vp_dev->msix_names), GFP_KERNEL);
	182	if (!vp_dev->msix_names)
	183	goto out_free_irq_vectors;
	184
	185	vp_dev->msix_affinity_masks = kcalloc(nvectors,
	186	sizeof(*vp_dev->msix_affinity_masks), GFP_KERNEL);
	187	if (!vp_dev->msix_affinity_masks)
	188	goto out_free_msix_names;
	189
	190	for (i = 0; i < nvectors; ++i) {
	191	if (!alloc_cpumask_var(&vp_dev->msix_affinity_masks[i],
	192	GFP_KERNEL))
	193	goto out_free_msix_affinity_masks;
	194	}
	195
	196	/* Set the vector used for configuration */
	197	snprintf(vp_dev->msix_names[0], sizeof(*vp_dev->msix_names),
	198	"%s-config", name);
	199	err = request_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_config_changed,
	200	0, vp_dev->msix_names[0], vp_dev);
a3cbec69	201	if (err)
52a61516	202	goto out_free_msix_affinity_masks;
a3cbec69	203
07ec5148 CH	204	/* Verify we had enough resources to assign the vector */
	205	if (vp_dev->config_vector(vp_dev, 0) == VIRTIO_MSI_NO_VECTOR) {
	206	err = -EBUSY;
	207	goto out_free_config_irq;
5c34d002 CH	208	}
5c34d002 CH	209
07ec5148 CH	210	vp_dev->msix_vector_map = kmalloc_array(nvqs,
	211	sizeof(*vp_dev->msix_vector_map), GFP_KERNEL);
	212	if (!vp_dev->msix_vector_map)
	213	goto out_disable_config_irq;
	214
	215	allocated_vectors = 1; /* vector 0 is the config interrupt */
d2a7ddda	216	for (i = 0; i < nvqs; ++i) {
6457f126 MT	217	if (!names[i]) {
	218	vqs[i] = NULL;
	219	continue;
a3cbec69 CH	220	}
a3cbec69 CH	221
07ec5148 CH	222	if (callbacks[i])
07ec5148 CH	223	msix_vec = allocated_vectors;
e969fed5	224	else
07ec5148 CH	225	msix_vec = VIRTIO_MSI_NO_VECTOR;
07ec5148 CH	226
5c34d002 CH	227	vqs[i] = vp_dev->setup_vq(vp_dev, i, callbacks[i], names[i],
5c34d002 CH	228	msix_vec);
e969fed5 MT	229	if (IS_ERR(vqs[i])) {
e969fed5 MT	230	err = PTR_ERR(vqs[i]);
07ec5148	231	goto out_remove_vqs;
e969fed5	232	}
0b22bd0b	233
5c34d002 CH	234	if (msix_vec == VIRTIO_MSI_NO_VECTOR) {
	235	vp_dev->msix_vector_map[i] = VIRTIO_MSI_NO_VECTOR;
	236	continue;
	237	}
	238
07ec5148 CH	239	snprintf(vp_dev->msix_names[i + 1],
07ec5148 CH	240	sizeof(*vp_dev->msix_names), "%s-%s",
0b22bd0b	241	dev_name(&vp_dev->vdev.dev), names[i]);
fa3a3279	242	err = request_irq(pci_irq_vector(vp_dev->pci_dev, msix_vec),
07ec5148 CH	243	vring_interrupt, IRQF_SHARED,
07ec5148 CH	244	vp_dev->msix_names[i + 1], vqs[i]);
5c34d002 CH	245	if (err) {
	246	/* don't free this irq on error */
	247	vp_dev->msix_vector_map[i] = VIRTIO_MSI_NO_VECTOR;
07ec5148	248	goto out_remove_vqs;
5c34d002 CH	249	}
5c34d002 CH	250	vp_dev->msix_vector_map[i] = msix_vec;
07ec5148	251
52a61516 CH	252	/*
	253	* Use a different vector for each queue if they are available,
	254	* else share the same vector for all VQs.
	255	*/
	256	if (!shared)
07ec5148	257	allocated_vectors++;
d2a7ddda	258	}
07ec5148	259
d2a7ddda MT	260	return 0;
d2a7ddda MT	261
07ec5148 CH	262	out_remove_vqs:
	263	vp_remove_vqs(vdev);
	264	kfree(vp_dev->msix_vector_map);
	265	out_disable_config_irq:
	266	vp_dev->config_vector(vp_dev, VIRTIO_MSI_NO_VECTOR);
	267	out_free_config_irq:
	268	free_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_dev);
	269	out_free_msix_affinity_masks:
	270	for (i = 0; i < nvectors; i++) {
	271	if (vp_dev->msix_affinity_masks[i])
	272	free_cpumask_var(vp_dev->msix_affinity_masks[i]);
	273	}
	274	kfree(vp_dev->msix_affinity_masks);
	275	out_free_msix_names:
	276	kfree(vp_dev->msix_names);
	277	out_free_irq_vectors:
	278	pci_free_irq_vectors(vp_dev->pci_dev);
e969fed5 MT	279	return err;
	280	}
	281
a3cbec69 CH	282	static int vp_find_vqs_intx(struct virtio_device *vdev, unsigned nvqs,
	283	struct virtqueue vqs[], vq_callback_t callbacks[],
	284	const char * const names[])
	285	{
	286	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	287	int i, err;
	288
a3cbec69 CH	289	err = request_irq(vp_dev->pci_dev->irq, vp_interrupt, IRQF_SHARED,
	290	dev_name(&vdev->dev), vp_dev);
	291	if (err)
07ec5148	292	return err;
a3cbec69	293
a3cbec69 CH	294	for (i = 0; i < nvqs; ++i) {
	295	if (!names[i]) {
	296	vqs[i] = NULL;
	297	continue;
	298	}
5c34d002	299	vqs[i] = vp_dev->setup_vq(vp_dev, i, callbacks[i], names[i],
a3cbec69 CH	300	VIRTIO_MSI_NO_VECTOR);
	301	if (IS_ERR(vqs[i])) {
	302	err = PTR_ERR(vqs[i]);
07ec5148	303	goto out_remove_vqs;
a3cbec69 CH	304	}
	305	}
	306
	307	return 0;
07ec5148 CH	308
	309	out_remove_vqs:
	310	vp_remove_vqs(vdev);
	311	free_irq(pci_irq_vector(vp_dev->pci_dev, 0), vp_dev);
a3cbec69 CH	312	return err;
	313	}
	314
e969fed5	315	/* the config->find_vqs() implementation */
38eb4a29	316	int vp_find_vqs(struct virtio_device *vdev, unsigned nvqs,
fb5e31d9 CH	317	struct virtqueue vqs[], vq_callback_t callbacks[],
fb5e31d9 CH	318	const char * const names[], struct irq_affinity *desc)
e969fed5	319	{
f68d2408	320	int err;
e969fed5	321
fb5e31d9	322	err = vp_find_vqs_msix(vdev, nvqs, vqs, callbacks, names, desc);
e969fed5 MT	323	if (!err)
e969fed5 MT	324	return 0;
a3cbec69	325	return vp_find_vqs_intx(vdev, nvqs, vqs, callbacks, names);
d2a7ddda MT	326	}
d2a7ddda MT	327
38eb4a29	328	const char vp_bus_name(struct virtio_device vdev)
66846048 RJ	329	{
	330	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	331
	332	return pci_name(vp_dev->pci_dev);
	333	}
	334
75a0a52b JW	335	/* Setup the affinity for a virtqueue:
	336	* - force the affinity for per vq vector
	337	* - OR over all affinities for shared MSI
	338	* - ignore the affinity request if we're using INTX
	339	*/
38eb4a29	340	int vp_set_vq_affinity(struct virtqueue *vq, int cpu)
75a0a52b JW	341	{
	342	struct virtio_device *vdev = vq->vdev;
	343	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
75a0a52b JW	344
	345	if (!vq->callback)
	346	return -EINVAL;
	347
53a020c6	348	if (vp_dev->pci_dev->msix_enabled) {
5c34d002 CH	349	int vec = vp_dev->msix_vector_map[vq->index];
	350	struct cpumask *mask = vp_dev->msix_affinity_masks[vec];
	351	unsigned int irq = pci_irq_vector(vp_dev->pci_dev, vec);
	352
75a0a52b JW	353	if (cpu == -1)
	354	irq_set_affinity_hint(irq, NULL);
	355	else {
210d150e	356	cpumask_clear(mask);
75a0a52b JW	357	cpumask_set_cpu(cpu, mask);
	358	irq_set_affinity_hint(irq, mask);
	359	}
	360	}
	361	return 0;
	362	}
	363
bbaba479 CH	364	const struct cpumask vp_get_vq_affinity(struct virtio_device vdev, int index)
	365	{
	366	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	367	unsigned int *map = vp_dev->msix_vector_map;
	368
	369	if (!map \|\| map[index] == VIRTIO_MSI_NO_VECTOR)
	370	return NULL;
	371
	372	return pci_irq_get_affinity(vp_dev->pci_dev, map[index]);
	373	}
	374
9e266ece	375	#ifdef CONFIG_PM_SLEEP
f0fe6f11 AS	376	static int virtio_pci_freeze(struct device *dev)
	377	{
	378	struct pci_dev *pci_dev = to_pci_dev(dev);
	379	struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
f0fe6f11 AS	380	int ret;
f0fe6f11 AS	381
c6716bae	382	ret = virtio_device_freeze(&vp_dev->vdev);
f0fe6f11 AS	383
	384	if (!ret)
	385	pci_disable_device(pci_dev);
	386	return ret;
	387	}
	388
0517fdd1	389	static int virtio_pci_restore(struct device *dev)
f0fe6f11 AS	390	{
	391	struct pci_dev *pci_dev = to_pci_dev(dev);
	392	struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
	393	int ret;
	394
	395	ret = pci_enable_device(pci_dev);
	396	if (ret)
	397	return ret;
0517fdd1	398
f0fe6f11	399	pci_set_master(pci_dev);
c6716bae	400	return virtio_device_restore(&vp_dev->vdev);
f0fe6f11 AS	401	}
f0fe6f11 AS	402
9a4253db	403	static const struct dev_pm_ops virtio_pci_pm_ops = {
f878d0be	404	SET_SYSTEM_SLEEP_PM_OPS(virtio_pci_freeze, virtio_pci_restore)
d0775363	405	};
3343660d	406	#endif
9a4253db MT	407
	408
	409	/* Qumranet donated their vendor ID for devices 0x1000 thru 0x10FF. */
	410	static const struct pci_device_id virtio_pci_id_table[] = {
caf02abf	411	{ PCI_DEVICE(PCI_VENDOR_ID_REDHAT_QUMRANET, PCI_ANY_ID) },
9a4253db MT	412	{ 0 }
	413	};
	414
	415	MODULE_DEVICE_TABLE(pci, virtio_pci_id_table);
	416
ff31d2e2 MT	417	static void virtio_pci_release_dev(struct device *_d)
	418	{
	419	struct virtio_device *vdev = dev_to_virtio(_d);
	420	struct virtio_pci_device *vp_dev = to_vp_device(vdev);
	421
	422	/* As struct device is a kobject, it's not safe to
	423	* free the memory (including the reference counter itself)
	424	* until it's release callback. */
	425	kfree(vp_dev);
	426	}
	427
9a4253db MT	428	static int virtio_pci_probe(struct pci_dev *pci_dev,
	429	const struct pci_device_id *id)
	430	{
ff31d2e2 MT	431	struct virtio_pci_device *vp_dev;
	432	int rc;
	433
	434	/* allocate our structure and fill it out */
	435	vp_dev = kzalloc(sizeof(struct virtio_pci_device), GFP_KERNEL);
	436	if (!vp_dev)
	437	return -ENOMEM;
	438
	439	pci_set_drvdata(pci_dev, vp_dev);
	440	vp_dev->vdev.dev.parent = &pci_dev->dev;
	441	vp_dev->vdev.dev.release = virtio_pci_release_dev;
	442	vp_dev->pci_dev = pci_dev;
ff31d2e2	443
ff31d2e2 MT	444	/* enable the device */
	445	rc = pci_enable_device(pci_dev);
	446	if (rc)
	447	goto err_enable_device;
	448
ac399d8f	449	if (force_legacy) {
1fcf0512	450	rc = virtio_pci_legacy_probe(vp_dev);
ac399d8f MT	451	/* Also try modern mode if we can't map BAR0 (no IO space). */
	452	if (rc == -ENODEV \|\| rc == -ENOMEM)
	453	rc = virtio_pci_modern_probe(vp_dev);
	454	if (rc)
	455	goto err_probe;
	456	} else {
	457	rc = virtio_pci_modern_probe(vp_dev);
	458	if (rc == -ENODEV)
	459	rc = virtio_pci_legacy_probe(vp_dev);
	460	if (rc)
	461	goto err_probe;
	462	}
ff31d2e2 MT	463
	464	pci_set_master(pci_dev);
	465
	466	rc = register_virtio_device(&vp_dev->vdev);
	467	if (rc)
	468	goto err_register;
	469
	470	return 0;
	471
	472	err_register:
1fcf0512 MT	473	if (vp_dev->ioaddr)
	474	virtio_pci_legacy_remove(vp_dev);
	475	else
	476	virtio_pci_modern_remove(vp_dev);
ff31d2e2	477	err_probe:
ff31d2e2 MT	478	pci_disable_device(pci_dev);
	479	err_enable_device:
	480	kfree(vp_dev);
	481	return rc;
9a4253db MT	482	}
	483
	484	static void virtio_pci_remove(struct pci_dev *pci_dev)
	485	{
ff31d2e2	486	struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
2989be09	487	struct device *dev = get_device(&vp_dev->vdev.dev);
ff31d2e2 MT	488
	489	unregister_virtio_device(&vp_dev->vdev);
	490
1fcf0512 MT	491	if (vp_dev->ioaddr)
	492	virtio_pci_legacy_remove(vp_dev);
	493	else
	494	virtio_pci_modern_remove(vp_dev);
ff31d2e2	495
ff31d2e2	496	pci_disable_device(pci_dev);
2989be09	497	put_device(dev);
9a4253db MT	498	}
	499
	500	static struct pci_driver virtio_pci_driver = {
	501	.name = "virtio-pci",
	502	.id_table = virtio_pci_id_table,
	503	.probe = virtio_pci_probe,
	504	.remove = virtio_pci_remove,
	505	#ifdef CONFIG_PM_SLEEP
	506	.driver.pm = &virtio_pci_pm_ops,
	507	#endif
	508	};
	509
	510	module_pci_driver(virtio_pci_driver);
5ff16110 HX	511
	512	MODULE_AUTHOR("Anthony Liguori <aliguori@us.ibm.com>");
	513	MODULE_DESCRIPTION("virtio-pci");
	514	MODULE_LICENSE("GPL");
	515	MODULE_VERSION("1");