irqfd

(Applies to kvm.git:66b0aed4, plus you will also need Davide Libenzi's eventfd_file_create() patch, which you can find here:

http://www.mail-archive.com/[email protected]/msg13923.html

You can find my complete tree with kvm.git, Davide's patch, and this series here:

http://git.kernel.org/?p=linux/kernel/git/ghaskins/vbus/linux-2.6.git;a=shortlog;h=irqfd

----------------------

irqfd, v5

This series implements a mechanism called "irqfd". It lets you create an eventfd based file-desriptor to inject interrupts to a kvm guest. For more details, please see the prologue for patch 2/2.

I am reasonably satisfied with this series, so Avi please consider for inclusion contingent on Davide formally submitting his eventfd_file_create() patch (and pending any new review comments, of course).

[ Changelog:

v5:
*) Added padding to the ioctl structure
*) Added proper ref-count increment to the file before returning
success. (Needs review by Al Viro, Davide Libenzi)
*) Cleaned up error-handling path to make sure we remove ourself
from the waitq if necessary.
*) Make sure we only add ourselves to kvm->irqfds if successful
creating the irqfd in the first place.
*) Rebased to kvm.git:66b0aed4

v4:
*) Changed allocation model to create the new fd last, after
we get past the last potential error point by using Davide's new eventfd_file_create interface (Al Viro, Davide Libenzi)
*) We no longer export sys_eventfd2() since it is replaced
functionally with eventfd_file_create();
*) Rebased to kvm.git:7da2e3ba

v3:
*) The kernel now allocates the eventfd (need to export sys_eventfd2) *) Added a flags field for future expansion to kvm_irqfd() *) We properly toggle the irq level 1+0.
*) We re-use the USERSPACE_SRC_ID instead of creating our own *) Properly check for failures establishing a poll-table with eventfd *) Fixed fd/file leaks on failure
*) Rebased to lateste kvm.git::41b76d8d04

v2:
*) Dropped notifier_chain based callbacks in favor of
wait_queue_t::func and file::poll based callbacks (Thanks to Davide for the suggestion)

v1:
*) Initial release

--------

We do not have a user of this interface in this series, though note future version of virtual-bus (v4 and above) will be based on this.

Note that this series requires userspace patches for qemu-kvm.git, v3, which you can find here: http://patchwork.kernel.org/patch/20213/

-Greg

---

Gregory Haskins (2):
kvm: add support for irqfd via eventfd-notification interface
eventfd: export eventfd interfaces for module use

 arch/x86/kvm/Makefile    |    2 -
 arch/x86/kvm/x86.c       |    1 
 fs/eventfd.c             |    4 +
 include/linux/kvm.h      |    8 ++
 include/linux/kvm_host.h |    4 +
 virt/kvm/irqfd.c         |  163 ++++++++++++++++++++++++++++++++++++++++++++++
 virt/kvm/kvm_main.c      |   11 +++
 7 files changed, 192 insertions(+), 1 deletions(-)
 create mode 100644 virt/kvm/irqfd.c

We will re-use eventfd for implmenting irqfd later in the series, and the irqfd users will potentially live in modules.

---

 fs/eventfd.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)

index 0de6ebb..2e1c2ff 100644
@@ -16,6 +16,7 @@
 #include <linux/anon_inodes.h>
 #include <linux/eventfd.h>
 #include <linux/syscalls.h>
+#include <linux/module.h>
 
 struct eventfd_ctx {
 	wait_queue_head_t wqh;

@@ -56,6 +57,7 @@ int eventfd_signal(struct file *file, int n)
 
 	return n;
 }
+EXPORT_SYMBOL_GPL(eventfd_signal);
 
 static int eventfd_release(struct inode *inode, struct file *file)
 {

@@ -197,6 +199,7 @@ struct file *eventfd_fget(int fd)
 
 	return file;
 }
+EXPORT_SYMBOL_GPL(eventfd_fget);
 
 struct file *eventfd_file_create(unsigned int count, int flags)
 {

@@ -225,6 +228,7 @@ struct file *eventfd_file_create(unsigned int count, int flags)
 
 	return file;
 }
+EXPORT_SYMBOL_GPL(eventfd_file_create);
 
 SYSCALL_DEFINE2(eventfd2, unsigned int, count, int, flags)
 {

KVM provides a complete virtual system environment for guests, including support for injecting interrupts modeled after the real exception/interrupt facilities present on the native platform (such as the IDT on x86). Virtual interrupts can come from a variety of sources (emulated devices, pass-through devices, etc) but all must be injected to the guest via the KVM infrastructure. This patch adds a new mechanism to inject a specific interrupt to a guest using a decoupled eventfd mechnanism: Any legal signal on the irqfd (using eventfd semantics from either userspace or kernel) will translate into an injected interrupt in the guest at the next available interrupt window.

---

 arch/x86/kvm/Makefile    |    2 -
 arch/x86/kvm/x86.c       |    1 
 include/linux/kvm.h      |    8 ++
 include/linux/kvm_host.h |    4 +
 virt/kvm/irqfd.c         |  163 ++++++++++++++++++++++++++++++++++++++++++++++
 virt/kvm/kvm_main.c      |   11 +++
 6 files changed, 188 insertions(+), 1 deletions(-)
 create mode 100644 virt/kvm/irqfd.c

index b43c4ef..d5fff51 100644
@@ -3,7 +3,7 @@
 #
 
 common-objs = $(addprefix ../../../virt/kvm/, kvm_main.o ioapic.o \
-                coalesced_mmio.o irq_comm.o)
+                coalesced_mmio.o irq_comm.o irqfd.o)
 ifeq ($(CONFIG_KVM_TRACE),y)
 common-objs += $(addprefix ../../../virt/kvm/, kvm_trace.o)
 endif

index 308d8e9..e68e415 100644
@@ -1026,6 +1026,7 @@ int kvm_dev_ioctl_check_extension(long ext)
 	case KVM_CAP_REINJECT_CONTROL:
 	case KVM_CAP_IRQ_INJECT_STATUS:
 	case KVM_CAP_ASSIGN_DEV_IRQ:
+	case KVM_CAP_IRQFD:
 		r = 1;
 		break;
 	case KVM_CAP_COALESCED_MMIO:

index 3db5d8d..ff8a08a 100644
@@ -415,6 +415,7 @@ struct kvm_trace_rec {
 #define KVM_CAP_ASSIGN_DEV_IRQ 29
 /* Another bug in KVM_SET_USER_MEMORY_REGION fixed: */
 #define KVM_CAP_JOIN_MEMORY_REGIONS_WORKS 30
+#define KVM_CAP_IRQFD 31
 
 #ifdef KVM_CAP_IRQ_ROUTING
 

@@ -454,6 +455,12 @@ struct kvm_irq_routing {
 
 #endif
 
+struct kvm_irqfd {
+	__u32 gsi;
+	__u32 flags;
+	__u8  pad[24];
+};
+
 /*
  * ioctls for VM fds
  */

@@ -498,6 +505,7 @@ struct kvm_irq_routing {
 #define KVM_ASSIGN_SET_MSIX_ENTRY \
 			_IOW(KVMIO, 0x74, struct kvm_assigned_msix_entry)
 #define KVM_DEASSIGN_DEV_IRQ       _IOW(KVMIO, 0x75, struct kvm_assigned_irq)
+#define KVM_IRQFD               _IOW(KVMIO, 0x76, struct kvm_irqfd)
 
 /*
  * ioctls for vcpu fds

index 095ebb6..6a8d1c1 100644
@@ -134,6 +134,7 @@ struct kvm {
 	struct list_head vm_list;
 	struct kvm_io_bus mmio_bus;
 	struct kvm_io_bus pio_bus;
+	struct list_head irqfds;
 	struct kvm_vm_stat stat;
 	struct kvm_arch arch;
 	atomic_t users_count;

@@ -524,4 +525,7 @@ static inline void kvm_free_irq_routing(struct kvm *kvm) {}
 
 #endif
 
+int kvm_irqfd(struct kvm *kvm, int gsi, int flags);
+void kvm_irqfd_release(struct kvm *kvm);
+
 #endif

index 2b73e19..8b3b06a 100644
@@ -972,6 +972,7 @@ static struct kvm *kvm_create_vm(void)
 	atomic_inc(&kvm->mm->mm_count);
 	spin_lock_init(&kvm->mmu_lock);
 	kvm_io_bus_init(&kvm->pio_bus);
+	INIT_LIST_HEAD(&kvm->irqfds);
 	mutex_init(&kvm->lock);
 	kvm_io_bus_init(&kvm->mmio_bus);
 	init_rwsem(&kvm->slots_lock);

@@ -1023,6 +1024,7 @@ static void kvm_destroy_vm(struct kvm *kvm)
 	spin_lock(&kvm_lock);
 	list_del(&kvm->vm_list);
 	spin_unlock(&kvm_lock);
+	kvm_irqfd_release(kvm);
 	kvm_free_irq_routing(kvm);
 	kvm_io_bus_destroy(&kvm->pio_bus);
 	kvm_io_bus_destroy(&kvm->mmio_bus);

@@ -2197,6 +2199,15 @@ static long kvm_vm_ioctl(struct file *filp,
 	}
 #endif
 #endif /* KVM_CAP_IRQ_ROUTING */
+	case KVM_IRQFD: {
+		struct kvm_irqfd data;
+
+		r = -EFAULT;
+		if (copy_from_user(&data, argp, sizeof data))
+			goto out;
+		r = kvm_irqfd(kvm, data.gsi, data.flags);
+		break;
+	}
 	default:
 		r = kvm_arch_vm_ioctl(filp, ioctl, arg);
 	}