linux/virt/kvm/eventfd.c

661 lines
15 KiB
C
Raw Normal View History

/*
* kvm eventfd support - use eventfd objects to signal various KVM events
*
* Copyright 2009 Novell. All Rights Reserved.
* Copyright 2010 Red Hat, Inc. and/or its affiliates.
*
* Author:
* Gregory Haskins <ghaskins@novell.com>
*
* This file is free software; you can redistribute it and/or modify
* it under the terms of version 2 of the GNU General Public License
* as published by the Free Software Foundation.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
*/
#include <linux/kvm_host.h>
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
#include <linux/kvm.h>
#include <linux/workqueue.h>
#include <linux/syscalls.h>
#include <linux/wait.h>
#include <linux/poll.h>
#include <linux/file.h>
#include <linux/list.h>
#include <linux/eventfd.h>
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
#include <linux/kernel.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 09:04:11 +01:00
#include <linux/slab.h>
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
#include "iodev.h"
/*
* --------------------------------------------------------------------
* irqfd: Allows an fd to be used to inject an interrupt to the guest
*
* Credit goes to Avi Kivity for the original idea.
* --------------------------------------------------------------------
*/
struct _irqfd {
/* Used for MSI fast-path */
struct kvm *kvm;
wait_queue_t wait;
/* Update side is protected by irqfds.lock */
struct kvm_kernel_irq_routing_entry __rcu *irq_entry;
/* Used for level IRQ fast-path */
int gsi;
struct work_struct inject;
/* Used for setup/shutdown */
struct eventfd_ctx *eventfd;
struct list_head list;
poll_table pt;
struct work_struct shutdown;
};
static struct workqueue_struct *irqfd_cleanup_wq;
static void
irqfd_inject(struct work_struct *work)
{
struct _irqfd *irqfd = container_of(work, struct _irqfd, inject);
struct kvm *kvm = irqfd->kvm;
kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID, irqfd->gsi, 1);
kvm_set_irq(kvm, KVM_USERSPACE_IRQ_SOURCE_ID, irqfd->gsi, 0);
}
/*
* Race-free decouple logic (ordering is critical)
*/
static void
irqfd_shutdown(struct work_struct *work)
{
struct _irqfd *irqfd = container_of(work, struct _irqfd, shutdown);
u64 cnt;
/*
* Synchronize with the wait-queue and unhook ourselves to prevent
* further events.
*/
eventfd_ctx_remove_wait_queue(irqfd->eventfd, &irqfd->wait, &cnt);
/*
* We know no new events will be scheduled at this point, so block
* until all previously outstanding events have completed
*/
flush_work_sync(&irqfd->inject);
/*
* It is now safe to release the object's resources
*/
eventfd_ctx_put(irqfd->eventfd);
kfree(irqfd);
}
/* assumes kvm->irqfds.lock is held */
static bool
irqfd_is_active(struct _irqfd *irqfd)
{
return list_empty(&irqfd->list) ? false : true;
}
/*
* Mark the irqfd as inactive and schedule it for removal
*
* assumes kvm->irqfds.lock is held
*/
static void
irqfd_deactivate(struct _irqfd *irqfd)
{
BUG_ON(!irqfd_is_active(irqfd));
list_del_init(&irqfd->list);
queue_work(irqfd_cleanup_wq, &irqfd->shutdown);
}
/*
* Called with wqh->lock held and interrupts disabled
*/
static int
irqfd_wakeup(wait_queue_t *wait, unsigned mode, int sync, void *key)
{
struct _irqfd *irqfd = container_of(wait, struct _irqfd, wait);
unsigned long flags = (unsigned long)key;
struct kvm_kernel_irq_routing_entry *irq;
struct kvm *kvm = irqfd->kvm;
if (flags & POLLIN) {
rcu_read_lock();
irq = rcu_dereference(irqfd->irq_entry);
/* An event has been signaled, inject an interrupt */
if (irq)
kvm_set_msi(irq, kvm, KVM_USERSPACE_IRQ_SOURCE_ID, 1);
else
schedule_work(&irqfd->inject);
rcu_read_unlock();
}
if (flags & POLLHUP) {
/* The eventfd is closing, detach from KVM */
unsigned long flags;
spin_lock_irqsave(&kvm->irqfds.lock, flags);
/*
* We must check if someone deactivated the irqfd before
* we could acquire the irqfds.lock since the item is
* deactivated from the KVM side before it is unhooked from
* the wait-queue. If it is already deactivated, we can
* simply return knowing the other side will cleanup for us.
* We cannot race against the irqfd going away since the
* other side is required to acquire wqh->lock, which we hold
*/
if (irqfd_is_active(irqfd))
irqfd_deactivate(irqfd);
spin_unlock_irqrestore(&kvm->irqfds.lock, flags);
}
return 0;
}
static void
irqfd_ptable_queue_proc(struct file *file, wait_queue_head_t *wqh,
poll_table *pt)
{
struct _irqfd *irqfd = container_of(pt, struct _irqfd, pt);
add_wait_queue(wqh, &irqfd->wait);
}
/* Must be called under irqfds.lock */
static void irqfd_update(struct kvm *kvm, struct _irqfd *irqfd,
struct kvm_irq_routing_table *irq_rt)
{
struct kvm_kernel_irq_routing_entry *e;
struct hlist_node *n;
if (irqfd->gsi >= irq_rt->nr_rt_entries) {
rcu_assign_pointer(irqfd->irq_entry, NULL);
return;
}
hlist_for_each_entry(e, n, &irq_rt->map[irqfd->gsi], link) {
/* Only fast-path MSI. */
if (e->type == KVM_IRQ_ROUTING_MSI)
rcu_assign_pointer(irqfd->irq_entry, e);
else
rcu_assign_pointer(irqfd->irq_entry, NULL);
}
}
static int
kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
{
struct kvm_irq_routing_table *irq_rt;
struct _irqfd *irqfd, *tmp;
struct file *file = NULL;
struct eventfd_ctx *eventfd = NULL;
int ret;
unsigned int events;
irqfd = kzalloc(sizeof(*irqfd), GFP_KERNEL);
if (!irqfd)
return -ENOMEM;
irqfd->kvm = kvm;
irqfd->gsi = args->gsi;
INIT_LIST_HEAD(&irqfd->list);
INIT_WORK(&irqfd->inject, irqfd_inject);
INIT_WORK(&irqfd->shutdown, irqfd_shutdown);
file = eventfd_fget(args->fd);
if (IS_ERR(file)) {
ret = PTR_ERR(file);
goto fail;
}
eventfd = eventfd_ctx_fileget(file);
if (IS_ERR(eventfd)) {
ret = PTR_ERR(eventfd);
goto fail;
}
irqfd->eventfd = eventfd;
/*
* Install our own custom wake-up handling so we are notified via
* a callback whenever someone signals the underlying eventfd
*/
init_waitqueue_func_entry(&irqfd->wait, irqfd_wakeup);
init_poll_funcptr(&irqfd->pt, irqfd_ptable_queue_proc);
spin_lock_irq(&kvm->irqfds.lock);
ret = 0;
list_for_each_entry(tmp, &kvm->irqfds.items, list) {
if (irqfd->eventfd != tmp->eventfd)
continue;
/* This fd is used for another irq already. */
ret = -EBUSY;
spin_unlock_irq(&kvm->irqfds.lock);
goto fail;
}
irq_rt = rcu_dereference_protected(kvm->irq_routing,
lockdep_is_held(&kvm->irqfds.lock));
irqfd_update(kvm, irqfd, irq_rt);
events = file->f_op->poll(file, &irqfd->pt);
list_add_tail(&irqfd->list, &kvm->irqfds.items);
/*
* Check if there was an event already pending on the eventfd
* before we registered, and trigger it as if we didn't miss it.
*/
if (events & POLLIN)
schedule_work(&irqfd->inject);
spin_unlock_irq(&kvm->irqfds.lock);
/*
* do not drop the file until the irqfd is fully initialized, otherwise
* we might race against the POLLHUP
*/
fput(file);
return 0;
fail:
if (eventfd && !IS_ERR(eventfd))
eventfd_ctx_put(eventfd);
if (!IS_ERR(file))
fput(file);
kfree(irqfd);
return ret;
}
void
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
kvm_eventfd_init(struct kvm *kvm)
{
spin_lock_init(&kvm->irqfds.lock);
INIT_LIST_HEAD(&kvm->irqfds.items);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
INIT_LIST_HEAD(&kvm->ioeventfds);
}
/*
* shutdown any irqfd's that match fd+gsi
*/
static int
kvm_irqfd_deassign(struct kvm *kvm, struct kvm_irqfd *args)
{
struct _irqfd *irqfd, *tmp;
struct eventfd_ctx *eventfd;
eventfd = eventfd_ctx_fdget(args->fd);
if (IS_ERR(eventfd))
return PTR_ERR(eventfd);
spin_lock_irq(&kvm->irqfds.lock);
list_for_each_entry_safe(irqfd, tmp, &kvm->irqfds.items, list) {
if (irqfd->eventfd == eventfd && irqfd->gsi == args->gsi) {
/*
* This rcu_assign_pointer is needed for when
* another thread calls kvm_irq_routing_update before
* we flush workqueue below (we synchronize with
* kvm_irq_routing_update using irqfds.lock).
* It is paired with synchronize_rcu done by caller
* of that function.
*/
rcu_assign_pointer(irqfd->irq_entry, NULL);
irqfd_deactivate(irqfd);
}
}
spin_unlock_irq(&kvm->irqfds.lock);
eventfd_ctx_put(eventfd);
/*
* Block until we know all outstanding shutdown jobs have completed
* so that we guarantee there will not be any more interrupts on this
* gsi once this deassign function returns.
*/
flush_workqueue(irqfd_cleanup_wq);
return 0;
}
int
kvm_irqfd(struct kvm *kvm, struct kvm_irqfd *args)
{
if (args->flags & ~KVM_IRQFD_FLAG_DEASSIGN)
return -EINVAL;
if (args->flags & KVM_IRQFD_FLAG_DEASSIGN)
return kvm_irqfd_deassign(kvm, args);
return kvm_irqfd_assign(kvm, args);
}
/*
* This function is called as the kvm VM fd is being released. Shutdown all
* irqfds that still remain open
*/
void
kvm_irqfd_release(struct kvm *kvm)
{
struct _irqfd *irqfd, *tmp;
spin_lock_irq(&kvm->irqfds.lock);
list_for_each_entry_safe(irqfd, tmp, &kvm->irqfds.items, list)
irqfd_deactivate(irqfd);
spin_unlock_irq(&kvm->irqfds.lock);
/*
* Block until we know all outstanding shutdown jobs have completed
* since we do not take a kvm* reference.
*/
flush_workqueue(irqfd_cleanup_wq);
}
/*
* Change irq_routing and irqfd.
* Caller must invoke synchronize_rcu afterwards.
*/
void kvm_irq_routing_update(struct kvm *kvm,
struct kvm_irq_routing_table *irq_rt)
{
struct _irqfd *irqfd;
spin_lock_irq(&kvm->irqfds.lock);
rcu_assign_pointer(kvm->irq_routing, irq_rt);
list_for_each_entry(irqfd, &kvm->irqfds.items, list)
irqfd_update(kvm, irqfd, irq_rt);
spin_unlock_irq(&kvm->irqfds.lock);
}
/*
* create a host-wide workqueue for issuing deferred shutdown requests
* aggregated from all vm* instances. We need our own isolated single-thread
* queue to prevent deadlock against flushing the normal work-queue.
*/
static int __init irqfd_module_init(void)
{
irqfd_cleanup_wq = create_singlethread_workqueue("kvm-irqfd-cleanup");
if (!irqfd_cleanup_wq)
return -ENOMEM;
return 0;
}
static void __exit irqfd_module_exit(void)
{
destroy_workqueue(irqfd_cleanup_wq);
}
module_init(irqfd_module_init);
module_exit(irqfd_module_exit);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
/*
* --------------------------------------------------------------------
* ioeventfd: translate a PIO/MMIO memory write to an eventfd signal.
*
* userspace can register a PIO/MMIO address with an eventfd for receiving
* notification when the memory has been touched.
* --------------------------------------------------------------------
*/
struct _ioeventfd {
struct list_head list;
u64 addr;
int length;
struct eventfd_ctx *eventfd;
u64 datamatch;
struct kvm_io_device dev;
bool wildcard;
};
static inline struct _ioeventfd *
to_ioeventfd(struct kvm_io_device *dev)
{
return container_of(dev, struct _ioeventfd, dev);
}
static void
ioeventfd_release(struct _ioeventfd *p)
{
eventfd_ctx_put(p->eventfd);
list_del(&p->list);
kfree(p);
}
static bool
ioeventfd_in_range(struct _ioeventfd *p, gpa_t addr, int len, const void *val)
{
u64 _val;
if (!(addr == p->addr && len == p->length))
/* address-range must be precise for a hit */
return false;
if (p->wildcard)
/* all else equal, wildcard is always a hit */
return true;
/* otherwise, we have to actually compare the data */
BUG_ON(!IS_ALIGNED((unsigned long)val, len));
switch (len) {
case 1:
_val = *(u8 *)val;
break;
case 2:
_val = *(u16 *)val;
break;
case 4:
_val = *(u32 *)val;
break;
case 8:
_val = *(u64 *)val;
break;
default:
return false;
}
return _val == p->datamatch ? true : false;
}
/* MMIO/PIO writes trigger an event if the addr/val match */
static int
ioeventfd_write(struct kvm_io_device *this, gpa_t addr, int len,
const void *val)
{
struct _ioeventfd *p = to_ioeventfd(this);
if (!ioeventfd_in_range(p, addr, len, val))
return -EOPNOTSUPP;
eventfd_signal(p->eventfd, 1);
return 0;
}
/*
* This function is called as KVM is completely shutting down. We do not
* need to worry about locking just nuke anything we have as quickly as possible
*/
static void
ioeventfd_destructor(struct kvm_io_device *this)
{
struct _ioeventfd *p = to_ioeventfd(this);
ioeventfd_release(p);
}
static const struct kvm_io_device_ops ioeventfd_ops = {
.write = ioeventfd_write,
.destructor = ioeventfd_destructor,
};
/* assumes kvm->slots_lock held */
static bool
ioeventfd_check_collision(struct kvm *kvm, struct _ioeventfd *p)
{
struct _ioeventfd *_p;
list_for_each_entry(_p, &kvm->ioeventfds, list)
if (_p->addr == p->addr && _p->length == p->length &&
(_p->wildcard || p->wildcard ||
_p->datamatch == p->datamatch))
return true;
return false;
}
static int
kvm_assign_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
{
int pio = args->flags & KVM_IOEVENTFD_FLAG_PIO;
enum kvm_bus bus_idx = pio ? KVM_PIO_BUS : KVM_MMIO_BUS;
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
struct _ioeventfd *p;
struct eventfd_ctx *eventfd;
int ret;
/* must be natural-word sized */
switch (args->len) {
case 1:
case 2:
case 4:
case 8:
break;
default:
return -EINVAL;
}
/* check for range overflow */
if (args->addr + args->len < args->addr)
return -EINVAL;
/* check for extra flags that we don't understand */
if (args->flags & ~KVM_IOEVENTFD_VALID_FLAG_MASK)
return -EINVAL;
eventfd = eventfd_ctx_fdget(args->fd);
if (IS_ERR(eventfd))
return PTR_ERR(eventfd);
p = kzalloc(sizeof(*p), GFP_KERNEL);
if (!p) {
ret = -ENOMEM;
goto fail;
}
INIT_LIST_HEAD(&p->list);
p->addr = args->addr;
p->length = args->len;
p->eventfd = eventfd;
/* The datamatch feature is optional, otherwise this is a wildcard */
if (args->flags & KVM_IOEVENTFD_FLAG_DATAMATCH)
p->datamatch = args->datamatch;
else
p->wildcard = true;
mutex_lock(&kvm->slots_lock);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
/* Verify that there isn't a match already */
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
if (ioeventfd_check_collision(kvm, p)) {
ret = -EEXIST;
goto unlock_fail;
}
kvm_iodevice_init(&p->dev, &ioeventfd_ops);
ret = kvm_io_bus_register_dev(kvm, bus_idx, p->addr, p->length,
&p->dev);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
if (ret < 0)
goto unlock_fail;
list_add_tail(&p->list, &kvm->ioeventfds);
mutex_unlock(&kvm->slots_lock);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
return 0;
unlock_fail:
mutex_unlock(&kvm->slots_lock);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
fail:
kfree(p);
eventfd_ctx_put(eventfd);
return ret;
}
static int
kvm_deassign_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
{
int pio = args->flags & KVM_IOEVENTFD_FLAG_PIO;
enum kvm_bus bus_idx = pio ? KVM_PIO_BUS : KVM_MMIO_BUS;
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
struct _ioeventfd *p, *tmp;
struct eventfd_ctx *eventfd;
int ret = -ENOENT;
eventfd = eventfd_ctx_fdget(args->fd);
if (IS_ERR(eventfd))
return PTR_ERR(eventfd);
mutex_lock(&kvm->slots_lock);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
list_for_each_entry_safe(p, tmp, &kvm->ioeventfds, list) {
bool wildcard = !(args->flags & KVM_IOEVENTFD_FLAG_DATAMATCH);
if (p->eventfd != eventfd ||
p->addr != args->addr ||
p->length != args->len ||
p->wildcard != wildcard)
continue;
if (!p->wildcard && p->datamatch != args->datamatch)
continue;
kvm_io_bus_unregister_dev(kvm, bus_idx, &p->dev);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
ioeventfd_release(p);
ret = 0;
break;
}
mutex_unlock(&kvm->slots_lock);
KVM: add ioeventfd support ioeventfd is a mechanism to register PIO/MMIO regions to trigger an eventfd signal when written to by a guest. Host userspace can register any arbitrary IO address with a corresponding eventfd and then pass the eventfd to a specific end-point of interest for handling. Normal IO requires a blocking round-trip since the operation may cause side-effects in the emulated model or may return data to the caller. Therefore, an IO in KVM traps from the guest to the host, causes a VMX/SVM "heavy-weight" exit back to userspace, and is ultimately serviced by qemu's device model synchronously before returning control back to the vcpu. However, there is a subclass of IO which acts purely as a trigger for other IO (such as to kick off an out-of-band DMA request, etc). For these patterns, the synchronous call is particularly expensive since we really only want to simply get our notification transmitted asychronously and return as quickly as possible. All the sychronous infrastructure to ensure proper data-dependencies are met in the normal IO case are just unecessary overhead for signalling. This adds additional computational load on the system, as well as latency to the signalling path. Therefore, we provide a mechanism for registration of an in-kernel trigger point that allows the VCPU to only require a very brief, lightweight exit just long enough to signal an eventfd. This also means that any clients compatible with the eventfd interface (which includes userspace and kernelspace equally well) can now register to be notified. The end result should be a more flexible and higher performance notification API for the backend KVM hypervisor and perhipheral components. To test this theory, we built a test-harness called "doorbell". This module has a function called "doorbell_ring()" which simply increments a counter for each time the doorbell is signaled. It supports signalling from either an eventfd, or an ioctl(). We then wired up two paths to the doorbell: One via QEMU via a registered io region and through the doorbell ioctl(). The other is direct via ioeventfd. You can download this test harness here: ftp://ftp.novell.com/dev/ghaskins/doorbell.tar.bz2 The measured results are as follows: qemu-mmio: 110000 iops, 9.09us rtt ioeventfd-mmio: 200100 iops, 5.00us rtt ioeventfd-pio: 367300 iops, 2.72us rtt I didn't measure qemu-pio, because I have to figure out how to register a PIO region with qemu's device model, and I got lazy. However, for now we can extrapolate based on the data from the NULLIO runs of +2.56us for MMIO, and -350ns for HC, we get: qemu-pio: 153139 iops, 6.53us rtt ioeventfd-hc: 412585 iops, 2.37us rtt these are just for fun, for now, until I can gather more data. Here is a graph for your convenience: http://developer.novell.com/wiki/images/7/76/Iofd-chart.png The conclusion to draw is that we save about 4us by skipping the userspace hop. -------------------- Signed-off-by: Gregory Haskins <ghaskins@novell.com> Acked-by: Michael S. Tsirkin <mst@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com>
2009-07-07 23:08:49 +02:00
eventfd_ctx_put(eventfd);
return ret;
}
int
kvm_ioeventfd(struct kvm *kvm, struct kvm_ioeventfd *args)
{
if (args->flags & KVM_IOEVENTFD_FLAG_DEASSIGN)
return kvm_deassign_ioeventfd(kvm, args);
return kvm_assign_ioeventfd(kvm, args);
}