2015-06-08 17:25:25 +02:00
|
|
|
/*
|
|
|
|
* vfio based device assignment support - platform devices
|
|
|
|
*
|
|
|
|
* Copyright Linaro Limited, 2014
|
|
|
|
*
|
|
|
|
* Authors:
|
|
|
|
* Kim Phillips <kim.phillips@linaro.org>
|
|
|
|
* Eric Auger <eric.auger@linaro.org>
|
|
|
|
*
|
|
|
|
* This work is licensed under the terms of the GNU GPL, version 2. See
|
|
|
|
* the COPYING file in the top-level directory.
|
|
|
|
*
|
|
|
|
* Based on vfio based PCI device assignment support:
|
|
|
|
* Copyright Red Hat, Inc. 2012
|
|
|
|
*/
|
|
|
|
|
2016-01-26 19:17:14 +01:00
|
|
|
#include "qemu/osdep.h"
|
include/qemu/osdep.h: Don't include qapi/error.h
Commit 57cb38b included qapi/error.h into qemu/osdep.h to get the
Error typedef. Since then, we've moved to include qemu/osdep.h
everywhere. Its file comment explains: "To avoid getting into
possible circular include dependencies, this file should not include
any other QEMU headers, with the exceptions of config-host.h,
compiler.h, os-posix.h and os-win32.h, all of which are doing a
similar job to this file and are under similar constraints."
qapi/error.h doesn't do a similar job, and it doesn't adhere to
similar constraints: it includes qapi-types.h. That's in excess of
100KiB of crap most .c files don't actually need.
Add the typedef to qemu/typedefs.h, and include that instead of
qapi/error.h. Include qapi/error.h in .c files that need it and don't
get it now. Include qapi-types.h in qom/object.h for uint16List.
Update scripts/clean-includes accordingly. Update it further to match
reality: replace config.h by config-target.h, add sysemu/os-posix.h,
sysemu/os-win32.h. Update the list of includes in the qemu/osdep.h
comment quoted above similarly.
This reduces the number of objects depending on qapi/error.h from "all
of them" to less than a third. Unfortunately, the number depending on
qapi-types.h shrinks only a little. More work is needed for that one.
Signed-off-by: Markus Armbruster <armbru@redhat.com>
[Fix compilation without the spice devel packages. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
2016-03-14 09:01:28 +01:00
|
|
|
#include "qapi/error.h"
|
2015-06-08 17:25:25 +02:00
|
|
|
#include <sys/ioctl.h>
|
2015-06-17 14:35:00 +02:00
|
|
|
#include <linux/vfio.h>
|
2015-06-08 17:25:25 +02:00
|
|
|
|
|
|
|
#include "hw/vfio/vfio-platform.h"
|
|
|
|
#include "qemu/error-report.h"
|
|
|
|
#include "qemu/range.h"
|
|
|
|
#include "sysemu/sysemu.h"
|
|
|
|
#include "exec/memory.h"
|
2018-05-29 01:26:59 +02:00
|
|
|
#include "exec/address-spaces.h"
|
2015-06-08 17:25:26 +02:00
|
|
|
#include "qemu/queue.h"
|
2015-06-08 17:25:25 +02:00
|
|
|
#include "hw/sysbus.h"
|
|
|
|
#include "trace.h"
|
|
|
|
#include "hw/platform-bus.h"
|
2015-07-06 20:15:14 +02:00
|
|
|
#include "sysemu/kvm.h"
|
2015-06-08 17:25:25 +02:00
|
|
|
|
2015-06-08 17:25:26 +02:00
|
|
|
/*
|
|
|
|
* Functions used whatever the injection method
|
|
|
|
*/
|
|
|
|
|
2015-10-05 20:30:12 +02:00
|
|
|
static inline bool vfio_irq_is_automasked(VFIOINTp *intp)
|
|
|
|
{
|
|
|
|
return intp->flags & VFIO_IRQ_INFO_AUTOMASKED;
|
|
|
|
}
|
|
|
|
|
2015-06-08 17:25:26 +02:00
|
|
|
/**
|
|
|
|
* vfio_init_intp - allocate, initialize the IRQ struct pointer
|
|
|
|
* and add it into the list of IRQs
|
|
|
|
* @vbasedev: the VFIO device handle
|
|
|
|
* @info: irq info struct retrieved from VFIO driver
|
2016-10-17 18:58:00 +02:00
|
|
|
* @errp: error object
|
2015-06-08 17:25:26 +02:00
|
|
|
*/
|
|
|
|
static VFIOINTp *vfio_init_intp(VFIODevice *vbasedev,
|
2016-10-17 18:58:00 +02:00
|
|
|
struct vfio_irq_info info, Error **errp)
|
2015-06-08 17:25:26 +02:00
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
VFIOPlatformDevice *vdev =
|
|
|
|
container_of(vbasedev, VFIOPlatformDevice, vbasedev);
|
|
|
|
SysBusDevice *sbdev = SYS_BUS_DEVICE(vdev);
|
|
|
|
VFIOINTp *intp;
|
|
|
|
|
|
|
|
intp = g_malloc0(sizeof(*intp));
|
|
|
|
intp->vdev = vdev;
|
|
|
|
intp->pin = info.index;
|
|
|
|
intp->flags = info.flags;
|
|
|
|
intp->state = VFIO_IRQ_INACTIVE;
|
2015-07-06 20:15:14 +02:00
|
|
|
intp->kvm_accel = false;
|
2015-06-08 17:25:26 +02:00
|
|
|
|
|
|
|
sysbus_init_irq(sbdev, &intp->qemuirq);
|
|
|
|
|
|
|
|
/* Get an eventfd for trigger */
|
2015-10-05 20:30:12 +02:00
|
|
|
intp->interrupt = g_malloc0(sizeof(EventNotifier));
|
|
|
|
ret = event_notifier_init(intp->interrupt, 0);
|
2015-06-08 17:25:26 +02:00
|
|
|
if (ret) {
|
2015-10-05 20:30:12 +02:00
|
|
|
g_free(intp->interrupt);
|
2015-06-08 17:25:26 +02:00
|
|
|
g_free(intp);
|
2016-10-17 18:58:00 +02:00
|
|
|
error_setg_errno(errp, -ret,
|
2019-05-21 17:15:42 +02:00
|
|
|
"failed to initialize trigger eventfd notifier");
|
2015-06-08 17:25:26 +02:00
|
|
|
return NULL;
|
|
|
|
}
|
2015-10-05 20:30:12 +02:00
|
|
|
if (vfio_irq_is_automasked(intp)) {
|
|
|
|
/* Get an eventfd for resample/unmask */
|
|
|
|
intp->unmask = g_malloc0(sizeof(EventNotifier));
|
|
|
|
ret = event_notifier_init(intp->unmask, 0);
|
|
|
|
if (ret) {
|
|
|
|
g_free(intp->interrupt);
|
|
|
|
g_free(intp->unmask);
|
|
|
|
g_free(intp);
|
2016-10-17 18:58:00 +02:00
|
|
|
error_setg_errno(errp, -ret,
|
2019-05-21 17:15:42 +02:00
|
|
|
"failed to initialize resample eventfd notifier");
|
2015-10-05 20:30:12 +02:00
|
|
|
return NULL;
|
|
|
|
}
|
2015-07-06 20:15:14 +02:00
|
|
|
}
|
2015-06-08 17:25:26 +02:00
|
|
|
|
|
|
|
QLIST_INSERT_HEAD(&vdev->intp_list, intp, next);
|
|
|
|
return intp;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_set_trigger_eventfd - set VFIO eventfd handling
|
|
|
|
*
|
|
|
|
* @intp: IRQ struct handle
|
|
|
|
* @handler: handler to be called on eventfd signaling
|
|
|
|
*
|
|
|
|
* Setup VFIO signaling and attach an optional user-side handler
|
|
|
|
* to the eventfd
|
|
|
|
*/
|
|
|
|
static int vfio_set_trigger_eventfd(VFIOINTp *intp,
|
|
|
|
eventfd_user_side_handler_t handler)
|
|
|
|
{
|
|
|
|
VFIODevice *vbasedev = &intp->vdev->vbasedev;
|
|
|
|
struct vfio_irq_set *irq_set;
|
|
|
|
int argsz, ret;
|
|
|
|
int32_t *pfd;
|
|
|
|
|
|
|
|
argsz = sizeof(*irq_set) + sizeof(*pfd);
|
|
|
|
irq_set = g_malloc0(argsz);
|
|
|
|
irq_set->argsz = argsz;
|
|
|
|
irq_set->flags = VFIO_IRQ_SET_DATA_EVENTFD | VFIO_IRQ_SET_ACTION_TRIGGER;
|
|
|
|
irq_set->index = intp->pin;
|
|
|
|
irq_set->start = 0;
|
|
|
|
irq_set->count = 1;
|
|
|
|
pfd = (int32_t *)&irq_set->data;
|
2015-10-05 20:30:12 +02:00
|
|
|
*pfd = event_notifier_get_fd(intp->interrupt);
|
2015-06-08 17:25:26 +02:00
|
|
|
qemu_set_fd_handler(*pfd, (IOHandler *)handler, NULL, intp);
|
|
|
|
ret = ioctl(vbasedev->fd, VFIO_DEVICE_SET_IRQS, irq_set);
|
|
|
|
if (ret < 0) {
|
|
|
|
error_report("vfio: Failed to set trigger eventfd: %m");
|
|
|
|
qemu_set_fd_handler(*pfd, NULL, NULL, NULL);
|
|
|
|
}
|
2017-07-26 19:38:17 +02:00
|
|
|
g_free(irq_set);
|
2015-06-08 17:25:26 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Functions only used when eventfds are handled on user-side
|
|
|
|
* ie. without irqfd
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_mmap_set_enabled - enable/disable the fast path mode
|
|
|
|
* @vdev: the VFIO platform device
|
|
|
|
* @enabled: the target mmap state
|
|
|
|
*
|
|
|
|
* enabled = true ~ fast path = MMIO region is mmaped (no KVM TRAP);
|
|
|
|
* enabled = false ~ slow path = MMIO region is trapped and region callbacks
|
|
|
|
* are called; slow path enables to trap the device IRQ status register reset
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void vfio_mmap_set_enabled(VFIOPlatformDevice *vdev, bool enabled)
|
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
|
|
|
for (i = 0; i < vdev->vbasedev.num_regions; i++) {
|
2016-03-10 17:39:07 +01:00
|
|
|
vfio_region_mmaps_set_enabled(vdev->regions[i], enabled);
|
2015-06-08 17:25:26 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_intp_mmap_enable - timer function, restores the fast path
|
|
|
|
* if there is no more active IRQ
|
|
|
|
* @opaque: actually points to the VFIO platform device
|
|
|
|
*
|
|
|
|
* Called on mmap timer timout, this function checks whether the
|
|
|
|
* IRQ is still active and if not, restores the fast path.
|
|
|
|
* by construction a single eventfd is handled at a time.
|
|
|
|
* if the IRQ is still active, the timer is re-programmed.
|
|
|
|
*/
|
|
|
|
static void vfio_intp_mmap_enable(void *opaque)
|
|
|
|
{
|
|
|
|
VFIOINTp *tmp;
|
|
|
|
VFIOPlatformDevice *vdev = (VFIOPlatformDevice *)opaque;
|
|
|
|
|
|
|
|
qemu_mutex_lock(&vdev->intp_mutex);
|
|
|
|
QLIST_FOREACH(tmp, &vdev->intp_list, next) {
|
|
|
|
if (tmp->state == VFIO_IRQ_ACTIVE) {
|
|
|
|
trace_vfio_platform_intp_mmap_enable(tmp->pin);
|
|
|
|
/* re-program the timer to check active status later */
|
|
|
|
timer_mod(vdev->mmap_timer,
|
|
|
|
qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
|
|
|
|
vdev->mmap_timeout);
|
|
|
|
qemu_mutex_unlock(&vdev->intp_mutex);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
vfio_mmap_set_enabled(vdev, true);
|
|
|
|
qemu_mutex_unlock(&vdev->intp_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_intp_inject_pending_lockheld - Injects a pending IRQ
|
|
|
|
* @opaque: opaque pointer, in practice the VFIOINTp handle
|
|
|
|
*
|
|
|
|
* The function is called on a previous IRQ completion, from
|
|
|
|
* vfio_platform_eoi, while the intp_mutex is locked.
|
|
|
|
* Also in such situation, the slow path already is set and
|
|
|
|
* the mmap timer was already programmed.
|
|
|
|
*/
|
|
|
|
static void vfio_intp_inject_pending_lockheld(VFIOINTp *intp)
|
|
|
|
{
|
|
|
|
trace_vfio_platform_intp_inject_pending_lockheld(intp->pin,
|
2015-10-05 20:30:12 +02:00
|
|
|
event_notifier_get_fd(intp->interrupt));
|
2015-06-08 17:25:26 +02:00
|
|
|
|
|
|
|
intp->state = VFIO_IRQ_ACTIVE;
|
|
|
|
|
|
|
|
/* trigger the virtual IRQ */
|
|
|
|
qemu_set_irq(intp->qemuirq, 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_intp_interrupt - The user-side eventfd handler
|
|
|
|
* @opaque: opaque pointer which in practice is the VFIOINTp handle
|
|
|
|
*
|
|
|
|
* the function is entered in event handler context:
|
|
|
|
* the vIRQ is injected into the guest if there is no other active
|
|
|
|
* or pending IRQ.
|
|
|
|
*/
|
|
|
|
static void vfio_intp_interrupt(VFIOINTp *intp)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
VFIOINTp *tmp;
|
|
|
|
VFIOPlatformDevice *vdev = intp->vdev;
|
|
|
|
bool delay_handling = false;
|
|
|
|
|
|
|
|
qemu_mutex_lock(&vdev->intp_mutex);
|
|
|
|
if (intp->state == VFIO_IRQ_INACTIVE) {
|
|
|
|
QLIST_FOREACH(tmp, &vdev->intp_list, next) {
|
|
|
|
if (tmp->state == VFIO_IRQ_ACTIVE ||
|
|
|
|
tmp->state == VFIO_IRQ_PENDING) {
|
|
|
|
delay_handling = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (delay_handling) {
|
|
|
|
/*
|
|
|
|
* the new IRQ gets a pending status and is pushed in
|
|
|
|
* the pending queue
|
|
|
|
*/
|
|
|
|
intp->state = VFIO_IRQ_PENDING;
|
|
|
|
trace_vfio_intp_interrupt_set_pending(intp->pin);
|
|
|
|
QSIMPLEQ_INSERT_TAIL(&vdev->pending_intp_queue,
|
|
|
|
intp, pqnext);
|
2015-10-05 20:30:12 +02:00
|
|
|
ret = event_notifier_test_and_clear(intp->interrupt);
|
2015-06-08 17:25:26 +02:00
|
|
|
qemu_mutex_unlock(&vdev->intp_mutex);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
trace_vfio_platform_intp_interrupt(intp->pin,
|
2015-10-05 20:30:12 +02:00
|
|
|
event_notifier_get_fd(intp->interrupt));
|
2015-06-08 17:25:26 +02:00
|
|
|
|
2015-10-05 20:30:12 +02:00
|
|
|
ret = event_notifier_test_and_clear(intp->interrupt);
|
2015-06-08 17:25:26 +02:00
|
|
|
if (!ret) {
|
2015-06-29 22:56:26 +02:00
|
|
|
error_report("Error when clearing fd=%d (ret = %d)",
|
2015-10-05 20:30:12 +02:00
|
|
|
event_notifier_get_fd(intp->interrupt), ret);
|
2015-06-08 17:25:26 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
intp->state = VFIO_IRQ_ACTIVE;
|
|
|
|
|
|
|
|
/* sets slow path */
|
|
|
|
vfio_mmap_set_enabled(vdev, false);
|
|
|
|
|
|
|
|
/* trigger the virtual IRQ */
|
|
|
|
qemu_set_irq(intp->qemuirq, 1);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Schedule the mmap timer which will restore fastpath when no IRQ
|
|
|
|
* is active anymore
|
|
|
|
*/
|
|
|
|
if (vdev->mmap_timeout) {
|
|
|
|
timer_mod(vdev->mmap_timer,
|
|
|
|
qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) +
|
|
|
|
vdev->mmap_timeout);
|
|
|
|
}
|
|
|
|
qemu_mutex_unlock(&vdev->intp_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_platform_eoi - IRQ completion routine
|
|
|
|
* @vbasedev: the VFIO device handle
|
|
|
|
*
|
|
|
|
* De-asserts the active virtual IRQ and unmasks the physical IRQ
|
|
|
|
* (effective for level sensitive IRQ auto-masked by the VFIO driver).
|
|
|
|
* Then it handles next pending IRQ if any.
|
|
|
|
* eoi function is called on the first access to any MMIO region
|
|
|
|
* after an IRQ was triggered, trapped since slow path was set.
|
|
|
|
* It is assumed this access corresponds to the IRQ status
|
|
|
|
* register reset. With such a mechanism, a single IRQ can be
|
|
|
|
* handled at a time since there is no way to know which IRQ
|
|
|
|
* was completed by the guest (we would need additional details
|
|
|
|
* about the IRQ status register mask).
|
|
|
|
*/
|
|
|
|
static void vfio_platform_eoi(VFIODevice *vbasedev)
|
|
|
|
{
|
|
|
|
VFIOINTp *intp;
|
|
|
|
VFIOPlatformDevice *vdev =
|
|
|
|
container_of(vbasedev, VFIOPlatformDevice, vbasedev);
|
|
|
|
|
|
|
|
qemu_mutex_lock(&vdev->intp_mutex);
|
|
|
|
QLIST_FOREACH(intp, &vdev->intp_list, next) {
|
|
|
|
if (intp->state == VFIO_IRQ_ACTIVE) {
|
|
|
|
trace_vfio_platform_eoi(intp->pin,
|
2015-10-05 20:30:12 +02:00
|
|
|
event_notifier_get_fd(intp->interrupt));
|
2015-06-08 17:25:26 +02:00
|
|
|
intp->state = VFIO_IRQ_INACTIVE;
|
|
|
|
|
|
|
|
/* deassert the virtual IRQ */
|
|
|
|
qemu_set_irq(intp->qemuirq, 0);
|
|
|
|
|
2015-10-05 20:30:12 +02:00
|
|
|
if (vfio_irq_is_automasked(intp)) {
|
2015-06-08 17:25:26 +02:00
|
|
|
/* unmasks the physical level-sensitive IRQ */
|
|
|
|
vfio_unmask_single_irqindex(vbasedev, intp->pin);
|
|
|
|
}
|
|
|
|
|
|
|
|
/* a single IRQ can be active at a time */
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
/* in case there are pending IRQs, handle the first one */
|
|
|
|
if (!QSIMPLEQ_EMPTY(&vdev->pending_intp_queue)) {
|
|
|
|
intp = QSIMPLEQ_FIRST(&vdev->pending_intp_queue);
|
|
|
|
vfio_intp_inject_pending_lockheld(intp);
|
|
|
|
QSIMPLEQ_REMOVE_HEAD(&vdev->pending_intp_queue, pqnext);
|
|
|
|
}
|
|
|
|
qemu_mutex_unlock(&vdev->intp_mutex);
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_start_eventfd_injection - starts the virtual IRQ injection using
|
|
|
|
* user-side handled eventfds
|
2015-10-05 20:30:12 +02:00
|
|
|
* @sbdev: the sysbus device handle
|
|
|
|
* @irq: the qemu irq handle
|
2015-06-08 17:25:26 +02:00
|
|
|
*/
|
|
|
|
|
2015-10-05 20:30:12 +02:00
|
|
|
static void vfio_start_eventfd_injection(SysBusDevice *sbdev, qemu_irq irq)
|
2015-06-08 17:25:26 +02:00
|
|
|
{
|
|
|
|
int ret;
|
2015-10-05 20:30:12 +02:00
|
|
|
VFIOPlatformDevice *vdev = VFIO_PLATFORM_DEVICE(sbdev);
|
|
|
|
VFIOINTp *intp;
|
|
|
|
|
|
|
|
QLIST_FOREACH(intp, &vdev->intp_list, next) {
|
|
|
|
if (intp->qemuirq == irq) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
assert(intp);
|
2015-06-08 17:25:26 +02:00
|
|
|
|
|
|
|
ret = vfio_set_trigger_eventfd(intp, vfio_intp_interrupt);
|
|
|
|
if (ret) {
|
2015-10-05 20:30:12 +02:00
|
|
|
error_report("vfio: failed to start eventfd signaling for IRQ %d: %m",
|
|
|
|
intp->pin);
|
|
|
|
abort();
|
2015-06-08 17:25:26 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-07-06 20:15:14 +02:00
|
|
|
/*
|
|
|
|
* Functions used for irqfd
|
|
|
|
*/
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_set_resample_eventfd - sets the resamplefd for an IRQ
|
|
|
|
* @intp: the IRQ struct handle
|
|
|
|
* programs the VFIO driver to unmask this IRQ when the
|
|
|
|
* intp->unmask eventfd is triggered
|
|
|
|
*/
|
|
|
|
static int vfio_set_resample_eventfd(VFIOINTp *intp)
|
|
|
|
{
|
|
|
|
VFIODevice *vbasedev = &intp->vdev->vbasedev;
|
|
|
|
struct vfio_irq_set *irq_set;
|
|
|
|
int argsz, ret;
|
|
|
|
int32_t *pfd;
|
|
|
|
|
|
|
|
argsz = sizeof(*irq_set) + sizeof(*pfd);
|
|
|
|
irq_set = g_malloc0(argsz);
|
|
|
|
irq_set->argsz = argsz;
|
|
|
|
irq_set->flags = VFIO_IRQ_SET_DATA_EVENTFD | VFIO_IRQ_SET_ACTION_UNMASK;
|
|
|
|
irq_set->index = intp->pin;
|
|
|
|
irq_set->start = 0;
|
|
|
|
irq_set->count = 1;
|
|
|
|
pfd = (int32_t *)&irq_set->data;
|
2015-10-05 20:30:12 +02:00
|
|
|
*pfd = event_notifier_get_fd(intp->unmask);
|
2015-07-06 20:15:14 +02:00
|
|
|
qemu_set_fd_handler(*pfd, NULL, NULL, NULL);
|
|
|
|
ret = ioctl(vbasedev->fd, VFIO_DEVICE_SET_IRQS, irq_set);
|
|
|
|
g_free(irq_set);
|
|
|
|
if (ret < 0) {
|
|
|
|
error_report("vfio: Failed to set resample eventfd: %m");
|
|
|
|
}
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-10-05 20:30:12 +02:00
|
|
|
/**
|
|
|
|
* vfio_start_irqfd_injection - starts the virtual IRQ injection using
|
|
|
|
* irqfd
|
|
|
|
*
|
|
|
|
* @sbdev: the sysbus device handle
|
|
|
|
* @irq: the qemu irq handle
|
|
|
|
*
|
|
|
|
* In case the irqfd setup fails, we fallback to userspace handled eventfd
|
|
|
|
*/
|
2015-07-06 20:15:14 +02:00
|
|
|
static void vfio_start_irqfd_injection(SysBusDevice *sbdev, qemu_irq irq)
|
|
|
|
{
|
|
|
|
VFIOPlatformDevice *vdev = VFIO_PLATFORM_DEVICE(sbdev);
|
|
|
|
VFIOINTp *intp;
|
|
|
|
|
|
|
|
if (!kvm_irqfds_enabled() || !kvm_resamplefds_enabled() ||
|
|
|
|
!vdev->irqfd_allowed) {
|
2015-10-05 20:30:12 +02:00
|
|
|
goto fail_irqfd;
|
2015-07-06 20:15:14 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
QLIST_FOREACH(intp, &vdev->intp_list, next) {
|
|
|
|
if (intp->qemuirq == irq) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
assert(intp);
|
|
|
|
|
2015-10-05 20:30:12 +02:00
|
|
|
if (kvm_irqchip_add_irqfd_notifier(kvm_state, intp->interrupt,
|
|
|
|
intp->unmask, irq) < 0) {
|
2015-07-06 20:15:14 +02:00
|
|
|
goto fail_irqfd;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (vfio_set_trigger_eventfd(intp, NULL) < 0) {
|
|
|
|
goto fail_vfio;
|
|
|
|
}
|
2015-10-05 20:30:12 +02:00
|
|
|
if (vfio_irq_is_automasked(intp)) {
|
|
|
|
if (vfio_set_resample_eventfd(intp) < 0) {
|
|
|
|
goto fail_vfio;
|
|
|
|
}
|
|
|
|
trace_vfio_platform_start_level_irqfd_injection(intp->pin,
|
|
|
|
event_notifier_get_fd(intp->interrupt),
|
|
|
|
event_notifier_get_fd(intp->unmask));
|
|
|
|
} else {
|
|
|
|
trace_vfio_platform_start_edge_irqfd_injection(intp->pin,
|
|
|
|
event_notifier_get_fd(intp->interrupt));
|
2015-07-06 20:15:14 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
intp->kvm_accel = true;
|
|
|
|
|
|
|
|
return;
|
|
|
|
fail_vfio:
|
2015-10-05 20:30:12 +02:00
|
|
|
kvm_irqchip_remove_irqfd_notifier(kvm_state, intp->interrupt, irq);
|
2015-10-05 20:30:12 +02:00
|
|
|
error_report("vfio: failed to start eventfd signaling for IRQ %d: %m",
|
|
|
|
intp->pin);
|
|
|
|
abort();
|
2015-07-06 20:15:14 +02:00
|
|
|
fail_irqfd:
|
2015-10-05 20:30:12 +02:00
|
|
|
vfio_start_eventfd_injection(sbdev, irq);
|
2015-07-06 20:15:14 +02:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2015-06-08 17:25:25 +02:00
|
|
|
/* VFIO skeleton */
|
|
|
|
|
|
|
|
static void vfio_platform_compute_needs_reset(VFIODevice *vbasedev)
|
|
|
|
{
|
|
|
|
vbasedev->needs_reset = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* not implemented yet */
|
|
|
|
static int vfio_platform_hot_reset_multi(VFIODevice *vbasedev)
|
|
|
|
{
|
|
|
|
return -1;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_populate_device - Allocate and populate MMIO region
|
2015-06-08 17:25:26 +02:00
|
|
|
* and IRQ structs according to driver returned information
|
2015-06-08 17:25:25 +02:00
|
|
|
* @vbasedev: the VFIO device handle
|
2016-10-17 18:58:00 +02:00
|
|
|
* @errp: error object
|
2015-06-08 17:25:25 +02:00
|
|
|
*
|
|
|
|
*/
|
2016-10-17 18:58:00 +02:00
|
|
|
static int vfio_populate_device(VFIODevice *vbasedev, Error **errp)
|
2015-06-08 17:25:25 +02:00
|
|
|
{
|
2015-06-08 17:25:26 +02:00
|
|
|
VFIOINTp *intp, *tmp;
|
2015-06-08 17:25:25 +02:00
|
|
|
int i, ret = -1;
|
|
|
|
VFIOPlatformDevice *vdev =
|
|
|
|
container_of(vbasedev, VFIOPlatformDevice, vbasedev);
|
|
|
|
|
|
|
|
if (!(vbasedev->flags & VFIO_DEVICE_FLAGS_PLATFORM)) {
|
2016-10-17 18:58:00 +02:00
|
|
|
error_setg(errp, "this isn't a platform device");
|
2015-06-08 17:25:25 +02:00
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-06-11 10:44:40 +02:00
|
|
|
vdev->regions = g_new0(VFIORegion *, vbasedev->num_regions);
|
2015-06-08 17:25:25 +02:00
|
|
|
|
|
|
|
for (i = 0; i < vbasedev->num_regions; i++) {
|
2016-03-10 17:39:07 +01:00
|
|
|
char *name = g_strdup_printf("VFIO %s region %d\n", vbasedev->name, i);
|
2015-06-08 17:25:25 +02:00
|
|
|
|
2015-11-10 20:11:08 +01:00
|
|
|
vdev->regions[i] = g_new0(VFIORegion, 1);
|
2016-03-10 17:39:07 +01:00
|
|
|
ret = vfio_region_setup(OBJECT(vdev), vbasedev,
|
|
|
|
vdev->regions[i], i, name);
|
|
|
|
g_free(name);
|
2015-06-08 17:25:25 +02:00
|
|
|
if (ret) {
|
2016-10-17 18:58:00 +02:00
|
|
|
error_setg_errno(errp, -ret, "failed to get region %d info", i);
|
2015-06-08 17:25:25 +02:00
|
|
|
goto reg_error;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-06-08 17:25:26 +02:00
|
|
|
vdev->mmap_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL,
|
|
|
|
vfio_intp_mmap_enable, vdev);
|
|
|
|
|
|
|
|
QSIMPLEQ_INIT(&vdev->pending_intp_queue);
|
|
|
|
|
|
|
|
for (i = 0; i < vbasedev->num_irqs; i++) {
|
|
|
|
struct vfio_irq_info irq = { .argsz = sizeof(irq) };
|
|
|
|
|
|
|
|
irq.index = i;
|
|
|
|
ret = ioctl(vbasedev->fd, VFIO_DEVICE_GET_IRQ_INFO, &irq);
|
|
|
|
if (ret) {
|
2016-10-17 18:58:00 +02:00
|
|
|
error_setg_errno(errp, -ret, "failed to get device irq info");
|
2015-06-08 17:25:26 +02:00
|
|
|
goto irq_err;
|
|
|
|
} else {
|
|
|
|
trace_vfio_platform_populate_interrupts(irq.index,
|
|
|
|
irq.count,
|
|
|
|
irq.flags);
|
2016-10-17 18:58:00 +02:00
|
|
|
intp = vfio_init_intp(vbasedev, irq, errp);
|
2015-06-08 17:25:26 +02:00
|
|
|
if (!intp) {
|
2016-10-17 18:58:00 +02:00
|
|
|
ret = -1;
|
2015-06-08 17:25:26 +02:00
|
|
|
goto irq_err;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2015-06-08 17:25:25 +02:00
|
|
|
return 0;
|
2015-06-08 17:25:26 +02:00
|
|
|
irq_err:
|
|
|
|
timer_del(vdev->mmap_timer);
|
|
|
|
QLIST_FOREACH_SAFE(intp, &vdev->intp_list, next, tmp) {
|
|
|
|
QLIST_REMOVE(intp, next);
|
|
|
|
g_free(intp);
|
|
|
|
}
|
2015-06-08 17:25:25 +02:00
|
|
|
reg_error:
|
|
|
|
for (i = 0; i < vbasedev->num_regions; i++) {
|
2016-03-10 17:39:07 +01:00
|
|
|
if (vdev->regions[i]) {
|
|
|
|
vfio_region_finalize(vdev->regions[i]);
|
|
|
|
}
|
2015-06-08 17:25:25 +02:00
|
|
|
g_free(vdev->regions[i]);
|
|
|
|
}
|
|
|
|
g_free(vdev->regions);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* specialized functions for VFIO Platform devices */
|
|
|
|
static VFIODeviceOps vfio_platform_ops = {
|
|
|
|
.vfio_compute_needs_reset = vfio_platform_compute_needs_reset,
|
|
|
|
.vfio_hot_reset_multi = vfio_platform_hot_reset_multi,
|
2015-06-08 17:25:26 +02:00
|
|
|
.vfio_eoi = vfio_platform_eoi,
|
2015-06-08 17:25:25 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_base_device_init - perform preliminary VFIO setup
|
|
|
|
* @vbasedev: the VFIO device handle
|
2016-10-17 18:58:01 +02:00
|
|
|
* @errp: error object
|
2015-06-08 17:25:25 +02:00
|
|
|
*
|
|
|
|
* Implement the VFIO command sequence that allows to discover
|
|
|
|
* assigned device resources: group extraction, device
|
|
|
|
* fd retrieval, resource query.
|
|
|
|
* Precondition: the device name must be initialized
|
|
|
|
*/
|
2016-10-17 18:58:01 +02:00
|
|
|
static int vfio_base_device_init(VFIODevice *vbasedev, Error **errp)
|
2015-06-08 17:25:25 +02:00
|
|
|
{
|
|
|
|
VFIOGroup *group;
|
|
|
|
VFIODevice *vbasedev_iter;
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
char *tmp, group_path[PATH_MAX], *group_name;
|
2015-06-08 17:25:25 +02:00
|
|
|
ssize_t len;
|
|
|
|
struct stat st;
|
|
|
|
int groupid;
|
|
|
|
int ret;
|
|
|
|
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
/* @sysfsdev takes precedence over @host */
|
|
|
|
if (vbasedev->sysfsdev) {
|
|
|
|
g_free(vbasedev->name);
|
2018-03-01 08:08:06 +01:00
|
|
|
vbasedev->name = g_path_get_basename(vbasedev->sysfsdev);
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
} else {
|
|
|
|
if (!vbasedev->name || strchr(vbasedev->name, '/')) {
|
2016-10-17 18:58:01 +02:00
|
|
|
error_setg(errp, "wrong host device name");
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
return -EINVAL;
|
|
|
|
}
|
2015-06-08 17:25:25 +02:00
|
|
|
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
vbasedev->sysfsdev = g_strdup_printf("/sys/bus/platform/devices/%s",
|
|
|
|
vbasedev->name);
|
|
|
|
}
|
2015-06-08 17:25:25 +02:00
|
|
|
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
if (stat(vbasedev->sysfsdev, &st) < 0) {
|
2016-10-17 18:58:01 +02:00
|
|
|
error_setg_errno(errp, errno,
|
|
|
|
"failed to get the sysfs host device file status");
|
2015-06-08 17:25:25 +02:00
|
|
|
return -errno;
|
|
|
|
}
|
|
|
|
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
tmp = g_strdup_printf("%s/iommu_group", vbasedev->sysfsdev);
|
|
|
|
len = readlink(tmp, group_path, sizeof(group_path));
|
|
|
|
g_free(tmp);
|
|
|
|
|
|
|
|
if (len < 0 || len >= sizeof(group_path)) {
|
2016-10-17 18:58:01 +02:00
|
|
|
ret = len < 0 ? -errno : -ENAMETOOLONG;
|
|
|
|
error_setg_errno(errp, -ret, "no iommu_group found");
|
|
|
|
return ret;
|
2015-06-08 17:25:25 +02:00
|
|
|
}
|
|
|
|
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
group_path[len] = 0;
|
2015-06-08 17:25:25 +02:00
|
|
|
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
group_name = basename(group_path);
|
2015-06-08 17:25:25 +02:00
|
|
|
if (sscanf(group_name, "%d", &groupid) != 1) {
|
2016-10-17 18:58:01 +02:00
|
|
|
error_setg_errno(errp, errno, "failed to read %s", group_path);
|
2015-06-08 17:25:25 +02:00
|
|
|
return -errno;
|
|
|
|
}
|
|
|
|
|
|
|
|
trace_vfio_platform_base_device_init(vbasedev->name, groupid);
|
|
|
|
|
2016-10-17 18:58:01 +02:00
|
|
|
group = vfio_get_group(groupid, &address_space_memory, errp);
|
2015-06-08 17:25:25 +02:00
|
|
|
if (!group) {
|
2016-10-17 18:58:01 +02:00
|
|
|
return -ENOENT;
|
2015-06-08 17:25:25 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
QLIST_FOREACH(vbasedev_iter, &group->device_list, next) {
|
|
|
|
if (strcmp(vbasedev_iter->name, vbasedev->name) == 0) {
|
2016-10-17 18:58:01 +02:00
|
|
|
error_setg(errp, "device is already attached");
|
2015-06-08 17:25:25 +02:00
|
|
|
vfio_put_group(group);
|
|
|
|
return -EBUSY;
|
|
|
|
}
|
|
|
|
}
|
2016-10-17 18:58:01 +02:00
|
|
|
ret = vfio_get_device(group, vbasedev->name, vbasedev, errp);
|
2015-06-08 17:25:25 +02:00
|
|
|
if (ret) {
|
|
|
|
vfio_put_group(group);
|
2016-10-17 18:58:01 +02:00
|
|
|
return ret;
|
2015-06-08 17:25:25 +02:00
|
|
|
}
|
|
|
|
|
2016-10-17 18:58:01 +02:00
|
|
|
ret = vfio_populate_device(vbasedev, errp);
|
2015-06-08 17:25:25 +02:00
|
|
|
if (ret) {
|
|
|
|
vfio_put_group(group);
|
|
|
|
}
|
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
/**
|
|
|
|
* vfio_platform_realize - the device realize function
|
|
|
|
* @dev: device state pointer
|
|
|
|
* @errp: error
|
|
|
|
*
|
|
|
|
* initialize the device, its memory regions and IRQ structures
|
|
|
|
* IRQ are started separately
|
|
|
|
*/
|
|
|
|
static void vfio_platform_realize(DeviceState *dev, Error **errp)
|
|
|
|
{
|
|
|
|
VFIOPlatformDevice *vdev = VFIO_PLATFORM_DEVICE(dev);
|
|
|
|
SysBusDevice *sbdev = SYS_BUS_DEVICE(dev);
|
|
|
|
VFIODevice *vbasedev = &vdev->vbasedev;
|
|
|
|
int i, ret;
|
|
|
|
|
|
|
|
vbasedev->type = VFIO_DEVICE_TYPE_PLATFORM;
|
2017-07-10 18:39:43 +02:00
|
|
|
vbasedev->dev = dev;
|
2015-06-08 17:25:25 +02:00
|
|
|
vbasedev->ops = &vfio_platform_ops;
|
|
|
|
|
2018-02-06 19:08:26 +01:00
|
|
|
qemu_mutex_init(&vdev->intp_mutex);
|
|
|
|
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
trace_vfio_platform_realize(vbasedev->sysfsdev ?
|
|
|
|
vbasedev->sysfsdev : vbasedev->name,
|
|
|
|
vdev->compat);
|
2015-06-08 17:25:25 +02:00
|
|
|
|
2016-10-17 18:58:01 +02:00
|
|
|
ret = vfio_base_device_init(vbasedev, errp);
|
2015-06-08 17:25:25 +02:00
|
|
|
if (ret) {
|
2016-10-17 18:58:01 +02:00
|
|
|
goto out;
|
2015-06-08 17:25:25 +02:00
|
|
|
}
|
|
|
|
|
2018-10-15 18:52:09 +02:00
|
|
|
if (!vdev->compat) {
|
|
|
|
GError *gerr = NULL;
|
|
|
|
gchar *contents;
|
|
|
|
gsize length;
|
|
|
|
char *path;
|
|
|
|
|
|
|
|
path = g_strdup_printf("%s/of_node/compatible", vbasedev->sysfsdev);
|
|
|
|
if (!g_file_get_contents(path, &contents, &length, &gerr)) {
|
|
|
|
error_setg(errp, "%s", gerr->message);
|
|
|
|
g_error_free(gerr);
|
|
|
|
g_free(path);
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
g_free(path);
|
|
|
|
vdev->compat = contents;
|
|
|
|
for (vdev->num_compat = 0; length; vdev->num_compat++) {
|
|
|
|
size_t skip = strlen(contents) + 1;
|
|
|
|
contents += skip;
|
|
|
|
length -= skip;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-06-08 17:25:25 +02:00
|
|
|
for (i = 0; i < vbasedev->num_regions; i++) {
|
2016-03-10 17:39:07 +01:00
|
|
|
if (vfio_region_mmap(vdev->regions[i])) {
|
2018-10-17 10:26:29 +02:00
|
|
|
warn_report("%s mmap unsupported, performance may be slow",
|
|
|
|
memory_region_name(vdev->regions[i]->mem));
|
2016-03-10 17:39:07 +01:00
|
|
|
}
|
|
|
|
sysbus_init_mmio(sbdev, vdev->regions[i]->mem);
|
2015-06-08 17:25:25 +02:00
|
|
|
}
|
2016-10-17 18:58:01 +02:00
|
|
|
out:
|
|
|
|
if (!ret) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (vdev->vbasedev.name) {
|
2018-10-17 10:26:30 +02:00
|
|
|
error_prepend(errp, VFIO_MSG_PREFIX, vdev->vbasedev.name);
|
2016-10-17 18:58:01 +02:00
|
|
|
} else {
|
|
|
|
error_prepend(errp, "vfio error: ");
|
|
|
|
}
|
2015-06-08 17:25:25 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static const VMStateDescription vfio_platform_vmstate = {
|
2019-05-21 17:15:41 +02:00
|
|
|
.name = "vfio-platform",
|
2015-06-08 17:25:25 +02:00
|
|
|
.unmigratable = 1,
|
|
|
|
};
|
|
|
|
|
|
|
|
static Property vfio_platform_dev_properties[] = {
|
|
|
|
DEFINE_PROP_STRING("host", VFIOPlatformDevice, vbasedev.name),
|
vfio: Add sysfsdev property for pci & platform
vfio-pci currently requires a host= parameter, which comes in the
form of a PCI address in [domain:]<bus:slot.function> notation. We
expect to find a matching entry in sysfs for that under
/sys/bus/pci/devices/. vfio-platform takes a similar approach, but
defines the host= parameter to be a string, which can be matched
directly under /sys/bus/platform/devices/. On the PCI side, we have
some interest in using vfio to expose vGPU devices. These are not
actual discrete PCI devices, so they don't have a compatible host PCI
bus address or a device link where QEMU wants to look for it. There's
also really no requirement that vfio can only be used to expose
physical devices, a new vfio bus and iommu driver could expose a
completely emulated device. To fit within the vfio framework, it
would need a kernel struct device and associated IOMMU group, but
those are easy constraints to manage.
To support such devices, which would include vGPUs, that honor the
VFIO PCI programming API, but are not necessarily backed by a unique
PCI address, add support for specifying any device in sysfs. The
vfio API already has support for probing the device type to ensure
compatibility with either vfio-pci or vfio-platform.
With this, a vfio-pci device could either be specified as:
-device vfio-pci,host=02:00.0
or
-device vfio-pci,sysfsdev=/sys/devices/pci0000:00/0000:00:1c.0/0000:02:00.0
or even
-device vfio-pci,sysfsdev=/sys/bus/pci/devices/0000:02:00.0
When vGPU support comes along, this might look something more like:
-device vfio-pci,sysfsdev=/sys/devices/virtual/intel-vgpu/vgpu0@0000:00:02.0
NB - This is only a made up example path
The same change is made for vfio-platform, specifying sysfsdev has
precedence over the old host option.
Tested-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
2016-03-10 17:39:07 +01:00
|
|
|
DEFINE_PROP_STRING("sysfsdev", VFIOPlatformDevice, vbasedev.sysfsdev),
|
2015-09-23 21:04:44 +02:00
|
|
|
DEFINE_PROP_BOOL("x-no-mmap", VFIOPlatformDevice, vbasedev.no_mmap, false),
|
2015-06-08 17:25:26 +02:00
|
|
|
DEFINE_PROP_UINT32("mmap-timeout-ms", VFIOPlatformDevice,
|
|
|
|
mmap_timeout, 1100),
|
2015-07-06 20:15:14 +02:00
|
|
|
DEFINE_PROP_BOOL("x-irqfd", VFIOPlatformDevice, irqfd_allowed, true),
|
2015-06-08 17:25:25 +02:00
|
|
|
DEFINE_PROP_END_OF_LIST(),
|
|
|
|
};
|
|
|
|
|
|
|
|
static void vfio_platform_class_init(ObjectClass *klass, void *data)
|
|
|
|
{
|
|
|
|
DeviceClass *dc = DEVICE_CLASS(klass);
|
2015-07-06 20:15:14 +02:00
|
|
|
SysBusDeviceClass *sbc = SYS_BUS_DEVICE_CLASS(klass);
|
2015-06-08 17:25:25 +02:00
|
|
|
|
|
|
|
dc->realize = vfio_platform_realize;
|
|
|
|
dc->props = vfio_platform_dev_properties;
|
|
|
|
dc->vmsd = &vfio_platform_vmstate;
|
|
|
|
dc->desc = "VFIO-based platform device assignment";
|
2015-07-06 20:15:14 +02:00
|
|
|
sbc->connect_irq_notifier = vfio_start_irqfd_injection;
|
2015-06-08 17:25:25 +02:00
|
|
|
set_bit(DEVICE_CATEGORY_MISC, dc->categories);
|
2018-10-15 18:52:09 +02:00
|
|
|
/* Supported by TYPE_VIRT_MACHINE */
|
|
|
|
dc->user_creatable = true;
|
2015-06-08 17:25:25 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static const TypeInfo vfio_platform_dev_info = {
|
|
|
|
.name = TYPE_VFIO_PLATFORM,
|
|
|
|
.parent = TYPE_SYS_BUS_DEVICE,
|
|
|
|
.instance_size = sizeof(VFIOPlatformDevice),
|
|
|
|
.class_init = vfio_platform_class_init,
|
|
|
|
.class_size = sizeof(VFIOPlatformDeviceClass),
|
|
|
|
};
|
|
|
|
|
|
|
|
static void register_vfio_platform_dev_type(void)
|
|
|
|
{
|
|
|
|
type_register_static(&vfio_platform_dev_info);
|
|
|
|
}
|
|
|
|
|
|
|
|
type_init(register_vfio_platform_dev_type)
|