[PATCH v7 0/4] irqfd support for arm/arm64

[PATCH v7 0/4] irqfd support for arm/arm64

[Eric Auger]

This patch series enables irqfd on arm and arm64.

Irqfd framework enables to inject a virtual IRQ into a guest upon an
eventfd trigger. User-side uses KVM_IRQFD VM ioctl to provide KVM with
a kvm_irqfd struct that associates a VM, an eventfd, a virtual IRQ number
(aka. the gsi). When an actor signals the eventfd (typically a VFIO
platform driver), the kvm irqfd subsystem injects the gsi into the VM.

Resamplefd also is supported for level sensitive interrupts, ie. the
user can provide another eventfd that is triggered when the completion
of the virtual IRQ (gsi) is detected by the GIC.

The gsi must correspond to a shared peripheral interrupt (SPI), ie the
GIC interrupt ID is gsi + 32.

The rationale behind not supporting PPI irqfd injection is that
any device using a PPI would be a private-to-the-CPU device (timer for
instance), so its state would have to be context-switched along with the
VCPU and would require in-kernel wiring anyhow. It is not a relevant use
case for irqfds.

This patch enables CONFIG_HAVE_KVM_EVENTFD and CONFIG_HAVE_KVM_IRQFD.

No IRQ routing table is used, enabling to remove CONFIG_HAVE_KVM_IRQCHIP

The ARM virtual interrupt controller, the VGIC, is dynamically
instantiated. The user-space may attempt to assign an irqfd before
the virtual interrupt controller is ready. For that reason a
check is added in the generic irqfd code to test whether the virtual
interrupt controller is ready.

This work was tested with Calxeda Midway xgmac main interrupt with
qemu-system-arm and QEMU VFIO platform device. Also irqfd was proven
functional on several vhost-net prototypes.

Available on ssh://git.linaro.org/people/eric.auger/linux.git
branch irqfd_v7_stdalone_official_release

v6 -> v7:
- kvm_arch_intc_initialized call introduced in first patch file
  related to kvm_arch_intc_initialized (generic part)
- add Christoffer's acks

v5 -> v6:
- take into account Christoffer's comments:
  - rename macro and function enabling to check the state of virtual
    interrupt controller (kvm_arch_intc_initialized)
  - kvm_arch_intc_initialized is declared in kvm_host.h whatever the
    archi support.
  - squash v5 patch files 3 & 4
  - KVM_CAP_IRQFD support depends on vgic_present
  - add Christoffer's Reviewed-by on last patch file

v4 -> v5:
- add the capability to check whether vgic is initialized when
  assigning an irqfd.  objective is to avoid injecting IRQ before
  this vgic is ready: this corresponds to new patch files 2, 3, 4.
- do not specifically handle early virtual IRQ injections in
  kvm_set_irq.  In case of injection when vgic is not yet ready,
  simply return an error.  User-space now has means to force vgic
  init and get notified if irqfd assign takes place too early.
- squash [PATCH v4 2/3] KVM: arm: add irqfd support and
         [PATCH v4 3/3] KVM: arm64: add irqfd support
- add Acked-by's in KVM: arm/arm64: unset CONFIG_HAVE_KVM_IRQCHIP
- some comment rewording in vgic

v3 -> v4:
- rebase on 3.18rc5
- vgic dynamic instantiation brought new challenges:
  handling of irqfd injection when vgic is not ready
- unset of CONFIG_HAVE_KVM_IRQCHIP in a separate patch
- add arm64 enable
- vgic.c style modifications according to Christoffer comments

v2 -> v3:
- removal of irq.h from eventfd.c put in a separate patch to increase
  visibility
- properly expose KVM_CAP_IRQFD capability in arm.c
- remove CONFIG_HAVE_KVM_IRQCHIP meaningfull only if irq_comm.c is used

v1 -> v2:
- rebase on 3.17rc1
- move of the dist unlock in process_maintenance
- remove of dist lock in __kvm_vgic_sync_hwstate
- rewording of the commit message (add resamplefd reference)
- remove irq.h

Eric Auger (4):
  KVM: arm/arm64: unset CONFIG_HAVE_KVM_IRQCHIP
  KVM: introduce kvm_arch_intc_initialized and use it in irqfd
  KVM: arm/arm64: implement kvm_arch_intc_initialized
  KVM: arm/arm64: add irqfd support

 Documentation/virtual/kvm/api.txt |  6 +++-
 arch/arm/include/asm/kvm_host.h   |  2 ++
 arch/arm/include/uapi/asm/kvm.h   |  3 ++
 arch/arm/kvm/Kconfig              |  4 +--
 arch/arm/kvm/Makefile             |  2 +-
 arch/arm/kvm/arm.c                | 10 +++++++
 arch/arm64/include/asm/kvm_host.h |  2 ++
 arch/arm64/include/uapi/asm/kvm.h |  3 ++
 arch/arm64/kvm/Kconfig            |  3 +-
 arch/arm64/kvm/Makefile           |  2 +-
 include/linux/kvm_host.h          | 14 +++++++++
 virt/kvm/arm/vgic.c               | 63 ++++++++++++++++++++++++++++++++++++---
 virt/kvm/eventfd.c                |  3 ++
 13 files changed, 107 insertions(+), 10 deletions(-)

-- 
1.9.1

[PATCH v7 1/4] KVM: arm/arm64: unset CONFIG_HAVE_KVM_IRQCHIP

[Eric Auger]

CONFIG_HAVE_KVM_IRQCHIP is needed to support IRQ routing (along
with irq_comm.c and irqchip.c usage). This is not the case for
arm/arm64 currently.

This patch unsets the flag for both arm and arm64.

Signed-off-by: Eric Auger <eric.auger@linaro.org>
Acked-by: Christoffer Dall <christoffer.dall@linaro.org>
Acked-by: Will Deacon <will.deacon@arm.com>
---
 arch/arm/kvm/Kconfig   | 2 --
 arch/arm64/kvm/Kconfig | 1 -
 2 files changed, 3 deletions(-)

diff --git a/arch/arm/kvm/Kconfig b/arch/arm/kvm/Kconfig
index 466bd29..9f581b1 100644
--- a/arch/arm/kvm/Kconfig
+++ b/arch/arm/kvm/Kconfig
@@ -55,7 +55,6 @@ config KVM_ARM_MAX_VCPUS
 config KVM_ARM_VGIC
 	bool "KVM support for Virtual GIC"
 	depends on KVM_ARM_HOST && OF
-	select HAVE_KVM_IRQCHIP
 	default y
 	---help---
 	  Adds support for a hardware assisted, in-kernel GIC emulation.
@@ -63,7 +62,6 @@ config KVM_ARM_VGIC
 config KVM_ARM_TIMER
 	bool "KVM support for Architected Timers"
 	depends on KVM_ARM_VGIC && ARM_ARCH_TIMER
-	select HAVE_KVM_IRQCHIP
 	default y
 	---help---
 	  Adds support for the Architected Timers in virtual machines
diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
index 8ba85e9..279e1a0 100644
--- a/arch/arm64/kvm/Kconfig
+++ b/arch/arm64/kvm/Kconfig
@@ -50,7 +50,6 @@ config KVM_ARM_MAX_VCPUS
 config KVM_ARM_VGIC
 	bool
 	depends on KVM_ARM_HOST && OF
-	select HAVE_KVM_IRQCHIP
 	---help---
 	  Adds support for a hardware assisted, in-kernel GIC emulation.
 
-- 
1.9.1

[PATCH v7 1/4] KVM: arm/arm64: unset CONFIG_HAVE_KVM_IRQCHIP

[André Przywara]

On 01/15/2015 02:47 PM, Eric Auger wrote:
> CONFIG_HAVE_KVM_IRQCHIP is needed to support IRQ routing (along
> with irq_comm.c and irqchip.c usage). This is not the case for
> arm/arm64 currently.
> 
> This patch unsets the flag for both arm and arm64.
> 
> Signed-off-by: Eric Auger <eric.auger@linaro.org>
> Acked-by: Christoffer Dall <christoffer.dall@linaro.org>
> Acked-by: Will Deacon <will.deacon@arm.com>

Given it's current usage in the kernel HAVE_KVM_IRQCHIP seems to be
grossly misnamed these days, but that's certainly not the duty of this
patch to solve. So:

Reviewed-by: Andre Przywara <andre.przywara@arm.com>

Cheers,
Andre.

> ---
>  arch/arm/kvm/Kconfig   | 2 --
>  arch/arm64/kvm/Kconfig | 1 -
>  2 files changed, 3 deletions(-)
> 
> diff --git a/arch/arm/kvm/Kconfig b/arch/arm/kvm/Kconfig
> index 466bd29..9f581b1 100644
> --- a/arch/arm/kvm/Kconfig
> +++ b/arch/arm/kvm/Kconfig
> @@ -55,7 +55,6 @@ config KVM_ARM_MAX_VCPUS
>  config KVM_ARM_VGIC
>  	bool "KVM support for Virtual GIC"
>  	depends on KVM_ARM_HOST && OF
> -	select HAVE_KVM_IRQCHIP
>  	default y
>  	---help---
>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
> @@ -63,7 +62,6 @@ config KVM_ARM_VGIC
>  config KVM_ARM_TIMER
>  	bool "KVM support for Architected Timers"
>  	depends on KVM_ARM_VGIC && ARM_ARCH_TIMER
> -	select HAVE_KVM_IRQCHIP
>  	default y
>  	---help---
>  	  Adds support for the Architected Timers in virtual machines
> diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
> index 8ba85e9..279e1a0 100644
> --- a/arch/arm64/kvm/Kconfig
> +++ b/arch/arm64/kvm/Kconfig
> @@ -50,7 +50,6 @@ config KVM_ARM_MAX_VCPUS
>  config KVM_ARM_VGIC
>  	bool
>  	depends on KVM_ARM_HOST && OF
> -	select HAVE_KVM_IRQCHIP
>  	---help---
>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
>  
> 

[PATCH v7 2/4] KVM: introduce kvm_arch_intc_initialized and use it in irqfd

[Eric Auger]

Introduce __KVM_HAVE_ARCH_INTC_INITIALIZED define and
associated kvm_arch_intc_initialized function. This latter
allows to test whether the virtual interrupt controller is initialized
and ready to accept virtual IRQ injection. On some architectures,
the virtual interrupt controller is dynamically instantiated, justifying
that kind of check.

The new function can now be used by irqfd to check whether the
virtual interrupt controller is ready on KVM_IRQFD request. If not,
KVM_IRQFD returns -EAGAIN.

Signed-off-by: Eric Auger <eric.auger@linaro.org>
Acked-by: Christoffer Dall <christoffer.dall@linaro.org>

---
v6 -> v7:
- From now on, kvm_irqfd_assign calls kvm_arch_intc_initialized
  (previously introduced in subsequent patch file).
- add Christoffer's ack

v5 -> v6:
- rename function name and macro
- add kvm_arch_intc_initialized declaration in case the archi defines
  the macro
---
 include/linux/kvm_host.h | 14 ++++++++++++++
 virt/kvm/eventfd.c       |  3 +++
 2 files changed, 17 insertions(+)

diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 26f1060..702036a 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -706,6 +706,20 @@ static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
 #endif
 }
 
+#ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
+/*
+ * returns trues if the virtual interrupt controller is initialized and
+ * ready to accept virtual IRQ. On some architectures the virtual interrupt
+ * controller is dynamically instantiated and this is not always true.
+ */
+bool kvm_arch_intc_initialized(struct kvm *kvm);
+#else
+static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
+{
+	return true;
+}
+#endif
+
 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
 void kvm_arch_destroy_vm(struct kvm *kvm);
 void kvm_arch_sync_events(struct kvm *kvm);
diff --git a/virt/kvm/eventfd.c b/virt/kvm/eventfd.c
index 148b239..fc5f43e 100644
--- a/virt/kvm/eventfd.c
+++ b/virt/kvm/eventfd.c
@@ -311,6 +311,9 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
 	unsigned int events;
 	int idx;
 
+	if (!kvm_arch_intc_initialized(kvm))
+		return -EAGAIN;
+
 	irqfd = kzalloc(sizeof(*irqfd), GFP_KERNEL);
 	if (!irqfd)
 		return -ENOMEM;
-- 
1.9.1

[PATCH v7 2/4] KVM: introduce kvm_arch_intc_initialized and use it in irqfd

[André Przywara]

Hi Eric,

On 01/15/2015 02:47 PM, Eric Auger wrote:
> Introduce __KVM_HAVE_ARCH_INTC_INITIALIZED define and
> associated kvm_arch_intc_initialized function. This latter
> allows to test whether the virtual interrupt controller is initialized
> and ready to accept virtual IRQ injection. On some architectures,
> the virtual interrupt controller is dynamically instantiated, justifying
> that kind of check.
> 
> The new function can now be used by irqfd to check whether the
> virtual interrupt controller is ready on KVM_IRQFD request. If not,
> KVM_IRQFD returns -EAGAIN.
> 
> Signed-off-by: Eric Auger <eric.auger@linaro.org>
> Acked-by: Christoffer Dall <christoffer.dall@linaro.org>
> 
> ---
> v6 -> v7:
> - From now on, kvm_irqfd_assign calls kvm_arch_intc_initialized
>   (previously introduced in subsequent patch file).
> - add Christoffer's ack
> 
> v5 -> v6:
> - rename function name and macro
> - add kvm_arch_intc_initialized declaration in case the archi defines
>   the macro
> ---
>  include/linux/kvm_host.h | 14 ++++++++++++++
>  virt/kvm/eventfd.c       |  3 +++
>  2 files changed, 17 insertions(+)
> 
> diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
> index 26f1060..702036a 100644
> --- a/include/linux/kvm_host.h
> +++ b/include/linux/kvm_host.h
> @@ -706,6 +706,20 @@ static inline wait_queue_head_t *kvm_arch_vcpu_wq(struct kvm_vcpu *vcpu)
>  #endif
>  }
>  
> +#ifdef __KVM_HAVE_ARCH_INTC_INITIALIZED
> +/*
> + * returns trues if the virtual interrupt controller is initialized and

Nit:          true

Other than that:
Reviewed-by: Andre Przywara <andre.przywara@arm.com>

Cheers,
Andre.


> + * ready to accept virtual IRQ. On some architectures the virtual interrupt
> + * controller is dynamically instantiated and this is not always true.
> + */
> +bool kvm_arch_intc_initialized(struct kvm *kvm);
> +#else
> +static inline bool kvm_arch_intc_initialized(struct kvm *kvm)
> +{
> +	return true;
> +}
> +#endif
> +
>  int kvm_arch_init_vm(struct kvm *kvm, unsigned long type);
>  void kvm_arch_destroy_vm(struct kvm *kvm);
>  void kvm_arch_sync_events(struct kvm *kvm);
> diff --git a/virt/kvm/eventfd.c b/virt/kvm/eventfd.c
> index 148b239..fc5f43e 100644
> --- a/virt/kvm/eventfd.c
> +++ b/virt/kvm/eventfd.c
> @@ -311,6 +311,9 @@ kvm_irqfd_assign(struct kvm *kvm, struct kvm_irqfd *args)
>  	unsigned int events;
>  	int idx;
>  
> +	if (!kvm_arch_intc_initialized(kvm))
> +		return -EAGAIN;
> +
>  	irqfd = kzalloc(sizeof(*irqfd), GFP_KERNEL);
>  	if (!irqfd)
>  		return -ENOMEM;
> 

[PATCH v7 3/4] KVM: arm/arm64: implement kvm_arch_intc_initialized

[Eric Auger]

On arm/arm64 the VGIC is dynamically instantiated and it is useful
to expose its state, especially for irqfd setup.

This patch defines __KVM_HAVE_ARCH_INTC_INITIALIZED and
implements kvm_arch_intc_initialized.

Signed-off-by: Eric Auger <eric.auger@linaro.org>
Acked-by: Christoffer Dall <christoffer.dall@linaro.org>

---

v6 -> v7:
- usage of the new function in irqfd is introduced in previous patch
  file.
- Changed the title and the commit message accordingly

v5 -> v6:
- remove kvm_arch_intc_initialized declaration from
  architecture specific kvm_host.h since it is now in
  generic kvm_host.h
- squash v5 patch files 3 & 4
---
 arch/arm/include/asm/kvm_host.h   | 2 ++
 arch/arm/kvm/arm.c                | 5 +++++
 arch/arm64/include/asm/kvm_host.h | 2 ++
 3 files changed, 9 insertions(+)

diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h
index 254e065..8f87259 100644
--- a/arch/arm/include/asm/kvm_host.h
+++ b/arch/arm/include/asm/kvm_host.h
@@ -27,6 +27,8 @@
 #include <asm/fpstate.h>
 #include <kvm/arm_arch_timer.h>
 
+#define __KVM_HAVE_ARCH_INTC_INITIALIZED
+
 #if defined(CONFIG_KVM_ARM_MAX_VCPUS)
 #define KVM_MAX_VCPUS CONFIG_KVM_ARM_MAX_VCPUS
 #else
diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
index 2d6d910..9c905b4 100644
--- a/arch/arm/kvm/arm.c
+++ b/arch/arm/kvm/arm.c
@@ -454,6 +454,11 @@ static int kvm_vcpu_first_run_init(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
+bool kvm_arch_intc_initialized(struct kvm *kvm)
+{
+	return vgic_initialized(kvm);
+}
+
 static void vcpu_pause(struct kvm_vcpu *vcpu)
 {
 	wait_queue_head_t *wq = kvm_arch_vcpu_wq(vcpu);
diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h
index 0b7dfdb..dfb6135 100644
--- a/arch/arm64/include/asm/kvm_host.h
+++ b/arch/arm64/include/asm/kvm_host.h
@@ -28,6 +28,8 @@
 #include <asm/kvm_asm.h>
 #include <asm/kvm_mmio.h>
 
+#define __KVM_HAVE_ARCH_INTC_INITIALIZED
+
 #if defined(CONFIG_KVM_ARM_MAX_VCPUS)
 #define KVM_MAX_VCPUS CONFIG_KVM_ARM_MAX_VCPUS
 #else
-- 
1.9.1

[PATCH v7 3/4] KVM: arm/arm64: implement kvm_arch_intc_initialized

[André Przywara]

On 01/15/2015 02:47 PM, Eric Auger wrote:
> On arm/arm64 the VGIC is dynamically instantiated and it is useful
> to expose its state, especially for irqfd setup.
> 
> This patch defines __KVM_HAVE_ARCH_INTC_INITIALIZED and
> implements kvm_arch_intc_initialized.
> 
> Signed-off-by: Eric Auger <eric.auger@linaro.org>
> Acked-by: Christoffer Dall <christoffer.dall@linaro.org>

Reviewed-by: Andre Przywara <andre.przywara@arm.com>

Cheers,
Andre.

[PATCH v7 4/4] KVM: arm/arm64: add irqfd support

[Eric Auger]

This patch enables irqfd on arm/arm64.

Both irqfd and resamplefd are supported. Injection is implemented
in vgic.c without routing.

This patch enables CONFIG_HAVE_KVM_EVENTFD and CONFIG_HAVE_KVM_IRQFD.

KVM_CAP_IRQFD is now advertised. KVM_CAP_IRQFD_RESAMPLE capability
automatically is advertised as soon as CONFIG_HAVE_KVM_IRQFD is set.

Irqfd injection is restricted to SPI. The rationale behind not
supporting PPI irqfd injection is that any device using a PPI would
be a private-to-the-CPU device (timer for instance), so its state
would have to be context-switched along with the VCPU and would
require in-kernel wiring anyhow. It is not a relevant use case for
irqfds.

Signed-off-by: Eric Auger <eric.auger@linaro.org>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>

---

v5 -> v6:
- KVM_CAP_IRQFD support depends on vgic_present
- add Christoffer's Reviewed-by

v4 -> v5:
- squash [PATCH v4 3/3] KVM: arm64: add irqfd support into this patch
- some rewording in Documentation/virtual/kvm/api.txt and in vgic
  vgic_process_maintenance unlock comment.
- move explanation of why not supporting PPI into commit message
- in case of injection before gic readiness, -ENODEV is returned. It is
  up to the user space to avoid this situation.

v3 -> v4:
- reword commit message
- explain why we unlock the distributor before calling kvm_notify_acked_irq
- rename is_assigned_irq into has_notifier
- change EOI and injection kvm_debug format string
- remove error local variable in kvm_set_irq
- Move HAVE_KVM_IRQCHIP unset in a separate patch
- handle case were the irqfd injection is attempted before the vgic is ready.
  in such a case the notifier, if any, is called immediatly
- use nr_irqs to test spi is within correct range

v2 -> v3:
- removal of irq.h from eventfd.c put in a separate patch to increase
  visibility
- properly expose KVM_CAP_IRQFD capability in arm.c
- remove CONFIG_HAVE_KVM_IRQCHIP meaningfull only if irq_comm.c is used

v1 -> v2:
- rebase on 3.17rc1
- move of the dist unlock in process_maintenance
- remove of dist lock in __kvm_vgic_sync_hwstate
- rewording of the commit message (add resamplefd reference)
- remove irq.h
---
 Documentation/virtual/kvm/api.txt |  6 +++-
 arch/arm/include/uapi/asm/kvm.h   |  3 ++
 arch/arm/kvm/Kconfig              |  2 ++
 arch/arm/kvm/Makefile             |  2 +-
 arch/arm/kvm/arm.c                |  5 ++++
 arch/arm64/include/uapi/asm/kvm.h |  3 ++
 arch/arm64/kvm/Kconfig            |  2 ++
 arch/arm64/kvm/Makefile           |  2 +-
 virt/kvm/arm/vgic.c               | 63 ++++++++++++++++++++++++++++++++++++---
 9 files changed, 81 insertions(+), 7 deletions(-)

diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
index 0007fef..5ed8088 100644
--- a/Documentation/virtual/kvm/api.txt
+++ b/Documentation/virtual/kvm/api.txt
@@ -2231,7 +2231,7 @@ into the hash PTE second double word).
 4.75 KVM_IRQFD
 
 Capability: KVM_CAP_IRQFD
-Architectures: x86 s390
+Architectures: x86 s390 arm arm64
 Type: vm ioctl
 Parameters: struct kvm_irqfd (in)
 Returns: 0 on success, -1 on error
@@ -2257,6 +2257,10 @@ Note that closing the resamplefd is not sufficient to disable the
 irqfd.  The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
 and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
 
+On ARM/ARM64, the gsi field in the kvm_irqfd struct specifies the Shared
+Peripheral Interrupt (SPI) index, such that the GIC interrupt ID is
+given by gsi + 32.
+
 4.76 KVM_PPC_ALLOCATE_HTAB
 
 Capability: KVM_CAP_PPC_ALLOC_HTAB
diff --git a/arch/arm/include/uapi/asm/kvm.h b/arch/arm/include/uapi/asm/kvm.h
index 09ee408..77547bb 100644
--- a/arch/arm/include/uapi/asm/kvm.h
+++ b/arch/arm/include/uapi/asm/kvm.h
@@ -196,6 +196,9 @@ struct kvm_arch_memory_slot {
 /* Highest supported SPI, from VGIC_NR_IRQS */
 #define KVM_ARM_IRQ_GIC_MAX		127
 
+/* One single KVM irqchip, ie. the VGIC */
+#define KVM_NR_IRQCHIPS          1
+
 /* PSCI interface */
 #define KVM_PSCI_FN_BASE		0x95c1ba5e
 #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
diff --git a/arch/arm/kvm/Kconfig b/arch/arm/kvm/Kconfig
index 9f581b1..e519a40 100644
--- a/arch/arm/kvm/Kconfig
+++ b/arch/arm/kvm/Kconfig
@@ -24,6 +24,7 @@ config KVM
 	select KVM_MMIO
 	select KVM_ARM_HOST
 	depends on ARM_VIRT_EXT && ARM_LPAE
+	select HAVE_KVM_EVENTFD
 	---help---
 	  Support hosting virtualized guest machines. You will also
 	  need to select one or more of the processor modules below.
@@ -55,6 +56,7 @@ config KVM_ARM_MAX_VCPUS
 config KVM_ARM_VGIC
 	bool "KVM support for Virtual GIC"
 	depends on KVM_ARM_HOST && OF
+	select HAVE_KVM_IRQFD
 	default y
 	---help---
 	  Adds support for a hardware assisted, in-kernel GIC emulation.
diff --git a/arch/arm/kvm/Makefile b/arch/arm/kvm/Makefile
index f7057ed..859db09 100644
--- a/arch/arm/kvm/Makefile
+++ b/arch/arm/kvm/Makefile
@@ -15,7 +15,7 @@ AFLAGS_init.o := -Wa,-march=armv7-a$(plus_virt)
 AFLAGS_interrupts.o := -Wa,-march=armv7-a$(plus_virt)
 
 KVM := ../../../virt/kvm
-kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
+kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
 
 obj-y += kvm-arm.o init.o interrupts.o
 obj-y += arm.o handle_exit.o guest.o mmu.o emulate.o reset.o
diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
index 9c905b4..8f8c1da 100644
--- a/arch/arm/kvm/arm.c
+++ b/arch/arm/kvm/arm.c
@@ -172,6 +172,11 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	case KVM_CAP_IRQCHIP:
 		r = vgic_present;
 		break;
+#ifdef CONFIG_HAVE_KVM_IRQFD
+	case KVM_CAP_IRQFD:
+		r = vgic_present;
+		break;
+#endif
 	case KVM_CAP_DEVICE_CTRL:
 	case KVM_CAP_USER_MEMORY:
 	case KVM_CAP_SYNC_MMU:
diff --git a/arch/arm64/include/uapi/asm/kvm.h b/arch/arm64/include/uapi/asm/kvm.h
index 8e38878..1ed4417 100644
--- a/arch/arm64/include/uapi/asm/kvm.h
+++ b/arch/arm64/include/uapi/asm/kvm.h
@@ -182,6 +182,9 @@ struct kvm_arch_memory_slot {
 /* Highest supported SPI, from VGIC_NR_IRQS */
 #define KVM_ARM_IRQ_GIC_MAX		127
 
+/* One single KVM irqchip, ie. the VGIC */
+#define KVM_NR_IRQCHIPS          1
+
 /* PSCI interface */
 #define KVM_PSCI_FN_BASE		0x95c1ba5e
 #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
index 279e1a0..09c25c2 100644
--- a/arch/arm64/kvm/Kconfig
+++ b/arch/arm64/kvm/Kconfig
@@ -26,6 +26,7 @@ config KVM
 	select KVM_ARM_HOST
 	select KVM_ARM_VGIC
 	select KVM_ARM_TIMER
+	select HAVE_KVM_EVENTFD
 	---help---
 	  Support hosting virtualized guest machines.
 
@@ -50,6 +51,7 @@ config KVM_ARM_MAX_VCPUS
 config KVM_ARM_VGIC
 	bool
 	depends on KVM_ARM_HOST && OF
+	select HAVE_KVM_IRQFD
 	---help---
 	  Adds support for a hardware assisted, in-kernel GIC emulation.
 
diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
index 32a0961..2e6b827 100644
--- a/arch/arm64/kvm/Makefile
+++ b/arch/arm64/kvm/Makefile
@@ -11,7 +11,7 @@ ARM=../../../arch/arm/kvm
 
 obj-$(CONFIG_KVM_ARM_HOST) += kvm.o
 
-kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
+kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
 kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/arm.o $(ARM)/mmu.o $(ARM)/mmio.o
 kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/psci.o $(ARM)/perf.o
 
diff --git a/virt/kvm/arm/vgic.c b/virt/kvm/arm/vgic.c
index 03affc7..d1a5c70 100644
--- a/virt/kvm/arm/vgic.c
+++ b/virt/kvm/arm/vgic.c
@@ -1447,7 +1447,10 @@ epilog:
 static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
 {
 	u32 status = vgic_get_interrupt_status(vcpu);
+	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
 	bool level_pending = false;
+	struct kvm *kvm = vcpu->kvm;
+	bool has_notifier;
 
 	kvm_debug("STATUS = %08x\n", status);
 
@@ -1464,6 +1467,7 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
 			struct vgic_lr vlr = vgic_get_lr(vcpu, lr);
 			WARN_ON(vgic_irq_is_edge(vcpu, vlr.irq));
 
+			spin_lock(&dist->lock);
 			vgic_irq_clear_queued(vcpu, vlr.irq);
 			WARN_ON(vlr.state & LR_STATE_MASK);
 			vlr.state = 0;
@@ -1482,6 +1486,24 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
 			 */
 			vgic_dist_irq_clear_soft_pend(vcpu, vlr.irq);
 
+			/*
+			 * kvm_notify_acked_irq calls kvm_set_irq()
+			 * to reset the IRQ level. Need to release the
+			 * lock for kvm_set_irq to grab it.
+			 */
+			spin_unlock(&dist->lock);
+
+			has_notifier = kvm_irq_has_notifier(kvm, 0,
+						vlr.irq - VGIC_NR_PRIVATE_IRQS);
+
+			if (has_notifier) {
+				kvm_debug("Guest EOIed vIRQ %d on CPU %d\n",
+					  vlr.irq, vcpu->vcpu_id);
+				kvm_notify_acked_irq(kvm, 0,
+						vlr.irq - VGIC_NR_PRIVATE_IRQS);
+			}
+			spin_lock(&dist->lock);
+
 			/* Any additional pending interrupt? */
 			if (vgic_dist_irq_get_level(vcpu, vlr.irq)) {
 				vgic_cpu_irq_set(vcpu, vlr.irq);
@@ -1491,6 +1513,8 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
 				vgic_cpu_irq_clear(vcpu, vlr.irq);
 			}
 
+			spin_unlock(&dist->lock);
+
 			/*
 			 * Despite being EOIed, the LR may not have
 			 * been marked as empty.
@@ -1555,14 +1579,10 @@ void kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
 
 void kvm_vgic_sync_hwstate(struct kvm_vcpu *vcpu)
 {
-	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
-
 	if (!irqchip_in_kernel(vcpu->kvm))
 		return;
 
-	spin_lock(&dist->lock);
 	__kvm_vgic_sync_hwstate(vcpu);
-	spin_unlock(&dist->lock);
 }
 
 int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)
@@ -2481,3 +2501,38 @@ out_free_irq:
 	free_percpu_irq(vgic->maint_irq, kvm_get_running_vcpus());
 	return ret;
 }
+
+int kvm_irq_map_gsi(struct kvm *kvm,
+		    struct kvm_kernel_irq_routing_entry *entries,
+		    int gsi)
+{
+	return gsi;
+}
+
+int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
+{
+	return pin;
+}
+
+int kvm_set_irq(struct kvm *kvm, int irq_source_id,
+		u32 irq, int level, bool line_status)
+{
+	unsigned int spi = irq + VGIC_NR_PRIVATE_IRQS;
+
+	kvm_debug("irqfd sets vIRQ %d to %d\n", irq, level);
+
+	if (likely(vgic_initialized(kvm))) {
+		if (spi > kvm->arch.vgic.nr_irqs)
+			return -EINVAL;
+		return kvm_vgic_inject_irq(kvm, 0, spi, level);
+	} else
+		return -ENODEV;
+}
+
+/* MSI not implemented yet */
+int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e,
+		struct kvm *kvm, int irq_source_id,
+		int level, bool line_status)
+{
+	return 0;
+}
-- 
1.9.1

[PATCH v7 4/4] KVM: arm/arm64: add irqfd support

[André Przywara]

Hi Eric,

On 01/15/2015 02:47 PM, Eric Auger wrote:
> This patch enables irqfd on arm/arm64.
> 
> Both irqfd and resamplefd are supported. Injection is implemented
> in vgic.c without routing.
> 
> This patch enables CONFIG_HAVE_KVM_EVENTFD and CONFIG_HAVE_KVM_IRQFD.
> 
> KVM_CAP_IRQFD is now advertised. KVM_CAP_IRQFD_RESAMPLE capability
> automatically is advertised as soon as CONFIG_HAVE_KVM_IRQFD is set.
> 
> Irqfd injection is restricted to SPI. The rationale behind not
> supporting PPI irqfd injection is that any device using a PPI would
> be a private-to-the-CPU device (timer for instance), so its state
> would have to be context-switched along with the VCPU and would
> require in-kernel wiring anyhow. It is not a relevant use case for
> irqfds.
> 
> Signed-off-by: Eric Auger <eric.auger@linaro.org>
> Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
> 
> ---
> 
> v5 -> v6:
> - KVM_CAP_IRQFD support depends on vgic_present
> - add Christoffer's Reviewed-by
> 
> v4 -> v5:
> - squash [PATCH v4 3/3] KVM: arm64: add irqfd support into this patch
> - some rewording in Documentation/virtual/kvm/api.txt and in vgic
>   vgic_process_maintenance unlock comment.
> - move explanation of why not supporting PPI into commit message
> - in case of injection before gic readiness, -ENODEV is returned. It is
>   up to the user space to avoid this situation.
> 
> v3 -> v4:
> - reword commit message
> - explain why we unlock the distributor before calling kvm_notify_acked_irq
> - rename is_assigned_irq into has_notifier
> - change EOI and injection kvm_debug format string
> - remove error local variable in kvm_set_irq
> - Move HAVE_KVM_IRQCHIP unset in a separate patch
> - handle case were the irqfd injection is attempted before the vgic is ready.
>   in such a case the notifier, if any, is called immediatly
> - use nr_irqs to test spi is within correct range
> 
> v2 -> v3:
> - removal of irq.h from eventfd.c put in a separate patch to increase
>   visibility
> - properly expose KVM_CAP_IRQFD capability in arm.c
> - remove CONFIG_HAVE_KVM_IRQCHIP meaningfull only if irq_comm.c is used
> 
> v1 -> v2:
> - rebase on 3.17rc1
> - move of the dist unlock in process_maintenance
> - remove of dist lock in __kvm_vgic_sync_hwstate
> - rewording of the commit message (add resamplefd reference)
> - remove irq.h
> ---
>  Documentation/virtual/kvm/api.txt |  6 +++-
>  arch/arm/include/uapi/asm/kvm.h   |  3 ++
>  arch/arm/kvm/Kconfig              |  2 ++
>  arch/arm/kvm/Makefile             |  2 +-
>  arch/arm/kvm/arm.c                |  5 ++++
>  arch/arm64/include/uapi/asm/kvm.h |  3 ++
>  arch/arm64/kvm/Kconfig            |  2 ++
>  arch/arm64/kvm/Makefile           |  2 +-
>  virt/kvm/arm/vgic.c               | 63 ++++++++++++++++++++++++++++++++++++---
>  9 files changed, 81 insertions(+), 7 deletions(-)
> 
> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
> index 0007fef..5ed8088 100644
> --- a/Documentation/virtual/kvm/api.txt
> +++ b/Documentation/virtual/kvm/api.txt
> @@ -2231,7 +2231,7 @@ into the hash PTE second double word).
>  4.75 KVM_IRQFD
>  
>  Capability: KVM_CAP_IRQFD
> -Architectures: x86 s390
> +Architectures: x86 s390 arm arm64
>  Type: vm ioctl
>  Parameters: struct kvm_irqfd (in)
>  Returns: 0 on success, -1 on error
> @@ -2257,6 +2257,10 @@ Note that closing the resamplefd is not sufficient to disable the
>  irqfd.  The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
>  and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
>  
> +On ARM/ARM64, the gsi field in the kvm_irqfd struct specifies the Shared
> +Peripheral Interrupt (SPI) index, such that the GIC interrupt ID is
> +given by gsi + 32.
> +
>  4.76 KVM_PPC_ALLOCATE_HTAB
>  
>  Capability: KVM_CAP_PPC_ALLOC_HTAB
> diff --git a/arch/arm/include/uapi/asm/kvm.h b/arch/arm/include/uapi/asm/kvm.h
> index 09ee408..77547bb 100644
> --- a/arch/arm/include/uapi/asm/kvm.h
> +++ b/arch/arm/include/uapi/asm/kvm.h
> @@ -196,6 +196,9 @@ struct kvm_arch_memory_slot {
>  /* Highest supported SPI, from VGIC_NR_IRQS */
>  #define KVM_ARM_IRQ_GIC_MAX		127
>  
> +/* One single KVM irqchip, ie. the VGIC */
> +#define KVM_NR_IRQCHIPS          1
> +
>  /* PSCI interface */
>  #define KVM_PSCI_FN_BASE		0x95c1ba5e
>  #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
> diff --git a/arch/arm/kvm/Kconfig b/arch/arm/kvm/Kconfig
> index 9f581b1..e519a40 100644
> --- a/arch/arm/kvm/Kconfig
> +++ b/arch/arm/kvm/Kconfig
> @@ -24,6 +24,7 @@ config KVM
>  	select KVM_MMIO
>  	select KVM_ARM_HOST
>  	depends on ARM_VIRT_EXT && ARM_LPAE
> +	select HAVE_KVM_EVENTFD
>  	---help---
>  	  Support hosting virtualized guest machines. You will also
>  	  need to select one or more of the processor modules below.
> @@ -55,6 +56,7 @@ config KVM_ARM_MAX_VCPUS
>  config KVM_ARM_VGIC
>  	bool "KVM support for Virtual GIC"
>  	depends on KVM_ARM_HOST && OF
> +	select HAVE_KVM_IRQFD
>  	default y
>  	---help---
>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
> diff --git a/arch/arm/kvm/Makefile b/arch/arm/kvm/Makefile
> index f7057ed..859db09 100644
> --- a/arch/arm/kvm/Makefile
> +++ b/arch/arm/kvm/Makefile
> @@ -15,7 +15,7 @@ AFLAGS_init.o := -Wa,-march=armv7-a$(plus_virt)
>  AFLAGS_interrupts.o := -Wa,-march=armv7-a$(plus_virt)
>  
>  KVM := ../../../virt/kvm
> -kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
> +kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
>  
>  obj-y += kvm-arm.o init.o interrupts.o
>  obj-y += arm.o handle_exit.o guest.o mmu.o emulate.o reset.o
> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
> index 9c905b4..8f8c1da 100644
> --- a/arch/arm/kvm/arm.c
> +++ b/arch/arm/kvm/arm.c
> @@ -172,6 +172,11 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>  	case KVM_CAP_IRQCHIP:
>  		r = vgic_present;
>  		break;
> +#ifdef CONFIG_HAVE_KVM_IRQFD
> +	case KVM_CAP_IRQFD:
> +		r = vgic_present;
> +		break;
> +#endif
>  	case KVM_CAP_DEVICE_CTRL:
>  	case KVM_CAP_USER_MEMORY:
>  	case KVM_CAP_SYNC_MMU:
> diff --git a/arch/arm64/include/uapi/asm/kvm.h b/arch/arm64/include/uapi/asm/kvm.h
> index 8e38878..1ed4417 100644
> --- a/arch/arm64/include/uapi/asm/kvm.h
> +++ b/arch/arm64/include/uapi/asm/kvm.h
> @@ -182,6 +182,9 @@ struct kvm_arch_memory_slot {
>  /* Highest supported SPI, from VGIC_NR_IRQS */
>  #define KVM_ARM_IRQ_GIC_MAX		127
>  
> +/* One single KVM irqchip, ie. the VGIC */
> +#define KVM_NR_IRQCHIPS          1
> +
>  /* PSCI interface */
>  #define KVM_PSCI_FN_BASE		0x95c1ba5e
>  #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
> diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
> index 279e1a0..09c25c2 100644
> --- a/arch/arm64/kvm/Kconfig
> +++ b/arch/arm64/kvm/Kconfig
> @@ -26,6 +26,7 @@ config KVM
>  	select KVM_ARM_HOST
>  	select KVM_ARM_VGIC
>  	select KVM_ARM_TIMER
> +	select HAVE_KVM_EVENTFD
>  	---help---
>  	  Support hosting virtualized guest machines.
>  
> @@ -50,6 +51,7 @@ config KVM_ARM_MAX_VCPUS
>  config KVM_ARM_VGIC
>  	bool
>  	depends on KVM_ARM_HOST && OF
> +	select HAVE_KVM_IRQFD
>  	---help---
>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
>  
> diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
> index 32a0961..2e6b827 100644
> --- a/arch/arm64/kvm/Makefile
> +++ b/arch/arm64/kvm/Makefile
> @@ -11,7 +11,7 @@ ARM=../../../arch/arm/kvm
>  
>  obj-$(CONFIG_KVM_ARM_HOST) += kvm.o
>  
> -kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
> +kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
>  kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/arm.o $(ARM)/mmu.o $(ARM)/mmio.o
>  kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/psci.o $(ARM)/perf.o
>  
> diff --git a/virt/kvm/arm/vgic.c b/virt/kvm/arm/vgic.c
> index 03affc7..d1a5c70 100644
> --- a/virt/kvm/arm/vgic.c
> +++ b/virt/kvm/arm/vgic.c
> @@ -1447,7 +1447,10 @@ epilog:
>  static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>  {
>  	u32 status = vgic_get_interrupt_status(vcpu);
> +	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
>  	bool level_pending = false;
> +	struct kvm *kvm = vcpu->kvm;
> +	bool has_notifier;
>  
>  	kvm_debug("STATUS = %08x\n", status);
>  
> @@ -1464,6 +1467,7 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>  			struct vgic_lr vlr = vgic_get_lr(vcpu, lr);
>  			WARN_ON(vgic_irq_is_edge(vcpu, vlr.irq));
>  
> +			spin_lock(&dist->lock);
>  			vgic_irq_clear_queued(vcpu, vlr.irq);
>  			WARN_ON(vlr.state & LR_STATE_MASK);
>  			vlr.state = 0;
> @@ -1482,6 +1486,24 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>  			 */
>  			vgic_dist_irq_clear_soft_pend(vcpu, vlr.irq);
>  
> +			/*
> +			 * kvm_notify_acked_irq calls kvm_set_irq()
> +			 * to reset the IRQ level. Need to release the
> +			 * lock for kvm_set_irq to grab it.
> +			 */
> +			spin_unlock(&dist->lock);
> +
> +			has_notifier = kvm_irq_has_notifier(kvm, 0,
> +						vlr.irq - VGIC_NR_PRIVATE_IRQS);
> +
> +			if (has_notifier) {

Could that be racy? Can the irqfd be de-assigned meanwhile? Or am I
seeing ghosts here? ;-)

> +				kvm_debug("Guest EOIed vIRQ %d on CPU %d\n",
> +					  vlr.irq, vcpu->vcpu_id);
> +				kvm_notify_acked_irq(kvm, 0,
> +						vlr.irq - VGIC_NR_PRIVATE_IRQS);
> +			}
> +			spin_lock(&dist->lock);
> +
>  			/* Any additional pending interrupt? */
>  			if (vgic_dist_irq_get_level(vcpu, vlr.irq)) {
>  				vgic_cpu_irq_set(vcpu, vlr.irq);
> @@ -1491,6 +1513,8 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>  				vgic_cpu_irq_clear(vcpu, vlr.irq);
>  			}
>  
> +			spin_unlock(&dist->lock);
> +
>  			/*
>  			 * Despite being EOIed, the LR may not have
>  			 * been marked as empty.
> @@ -1555,14 +1579,10 @@ void kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
>  
>  void kvm_vgic_sync_hwstate(struct kvm_vcpu *vcpu)
>  {
> -	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
> -
>  	if (!irqchip_in_kernel(vcpu->kvm))
>  		return;
>  
> -	spin_lock(&dist->lock);
>  	__kvm_vgic_sync_hwstate(vcpu);
> -	spin_unlock(&dist->lock);
>  }

Given that you drop the coarse grained locking around
__kvm_vgic_sync_hwstate(), can you adjust the comment on top of it and
either delete it or relax the locking requirement?

Have I got this right that most of the locking is not needed as long as
we deal only with VCPU specific data structures, as any preemption would
not execute code handling this very VCPU again?


Also it would be beneficial to note the change of the locking in the
commit message (or even make it a separate patch). That would help
possible bisecting.

>  
>  int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)
> @@ -2481,3 +2501,38 @@ out_free_irq:
>  	free_percpu_irq(vgic->maint_irq, kvm_get_running_vcpus());
>  	return ret;
>  }
> +
> +int kvm_irq_map_gsi(struct kvm *kvm,
> +		    struct kvm_kernel_irq_routing_entry *entries,
> +		    int gsi)
> +{
> +	return gsi;
> +}
> +
> +int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
> +{
> +	return pin;
> +}
> +
> +int kvm_set_irq(struct kvm *kvm, int irq_source_id,
> +		u32 irq, int level, bool line_status)
> +{
> +	unsigned int spi = irq + VGIC_NR_PRIVATE_IRQS;
> +
> +	kvm_debug("irqfd sets vIRQ %d to %d\n", irq, level);
> +
> +	if (likely(vgic_initialized(kvm))) {

Is that really needed? I reckon we can only get here if an irqfd has
been assigned, which is guarded already by that initialization check.

> +		if (spi > kvm->arch.vgic.nr_irqs)

I wonder if that check could be done earlier at assignment time. Does
that not work because the assignment is generic code and has no notion
of the maximum supported IRQ numbers as the VGIC has?

Cheers,
Andre.

> +			return -EINVAL;
> +		return kvm_vgic_inject_irq(kvm, 0, spi, level);
> +	} else
> +		return -ENODEV;
> +}
> +
> +/* MSI not implemented yet */
> +int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e,
> +		struct kvm *kvm, int irq_source_id,
> +		int level, bool line_status)
> +{
> +	return 0;
> +}
> 

[PATCH v7 4/4] KVM: arm/arm64: add irqfd support

[Eric Auger]

Hi Andre,
On 01/16/2015 11:07 AM, Andr? Przywara wrote:
> Hi Eric,
> 
> On 01/15/2015 02:47 PM, Eric Auger wrote:
>> This patch enables irqfd on arm/arm64.
>>
>> Both irqfd and resamplefd are supported. Injection is implemented
>> in vgic.c without routing.
>>
>> This patch enables CONFIG_HAVE_KVM_EVENTFD and CONFIG_HAVE_KVM_IRQFD.
>>
>> KVM_CAP_IRQFD is now advertised. KVM_CAP_IRQFD_RESAMPLE capability
>> automatically is advertised as soon as CONFIG_HAVE_KVM_IRQFD is set.
>>
>> Irqfd injection is restricted to SPI. The rationale behind not
>> supporting PPI irqfd injection is that any device using a PPI would
>> be a private-to-the-CPU device (timer for instance), so its state
>> would have to be context-switched along with the VCPU and would
>> require in-kernel wiring anyhow. It is not a relevant use case for
>> irqfds.
>>
>> Signed-off-by: Eric Auger <eric.auger@linaro.org>
>> Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
>>
>> ---
>>
>> v5 -> v6:
>> - KVM_CAP_IRQFD support depends on vgic_present
>> - add Christoffer's Reviewed-by
>>
>> v4 -> v5:
>> - squash [PATCH v4 3/3] KVM: arm64: add irqfd support into this patch
>> - some rewording in Documentation/virtual/kvm/api.txt and in vgic
>>   vgic_process_maintenance unlock comment.
>> - move explanation of why not supporting PPI into commit message
>> - in case of injection before gic readiness, -ENODEV is returned. It is
>>   up to the user space to avoid this situation.
>>
>> v3 -> v4:
>> - reword commit message
>> - explain why we unlock the distributor before calling kvm_notify_acked_irq
>> - rename is_assigned_irq into has_notifier
>> - change EOI and injection kvm_debug format string
>> - remove error local variable in kvm_set_irq
>> - Move HAVE_KVM_IRQCHIP unset in a separate patch
>> - handle case were the irqfd injection is attempted before the vgic is ready.
>>   in such a case the notifier, if any, is called immediatly
>> - use nr_irqs to test spi is within correct range
>>
>> v2 -> v3:
>> - removal of irq.h from eventfd.c put in a separate patch to increase
>>   visibility
>> - properly expose KVM_CAP_IRQFD capability in arm.c
>> - remove CONFIG_HAVE_KVM_IRQCHIP meaningfull only if irq_comm.c is used
>>
>> v1 -> v2:
>> - rebase on 3.17rc1
>> - move of the dist unlock in process_maintenance
>> - remove of dist lock in __kvm_vgic_sync_hwstate
>> - rewording of the commit message (add resamplefd reference)
>> - remove irq.h
>> ---
>>  Documentation/virtual/kvm/api.txt |  6 +++-
>>  arch/arm/include/uapi/asm/kvm.h   |  3 ++
>>  arch/arm/kvm/Kconfig              |  2 ++
>>  arch/arm/kvm/Makefile             |  2 +-
>>  arch/arm/kvm/arm.c                |  5 ++++
>>  arch/arm64/include/uapi/asm/kvm.h |  3 ++
>>  arch/arm64/kvm/Kconfig            |  2 ++
>>  arch/arm64/kvm/Makefile           |  2 +-
>>  virt/kvm/arm/vgic.c               | 63 ++++++++++++++++++++++++++++++++++++---
>>  9 files changed, 81 insertions(+), 7 deletions(-)
>>
>> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
>> index 0007fef..5ed8088 100644
>> --- a/Documentation/virtual/kvm/api.txt
>> +++ b/Documentation/virtual/kvm/api.txt
>> @@ -2231,7 +2231,7 @@ into the hash PTE second double word).
>>  4.75 KVM_IRQFD
>>  
>>  Capability: KVM_CAP_IRQFD
>> -Architectures: x86 s390
>> +Architectures: x86 s390 arm arm64
>>  Type: vm ioctl
>>  Parameters: struct kvm_irqfd (in)
>>  Returns: 0 on success, -1 on error
>> @@ -2257,6 +2257,10 @@ Note that closing the resamplefd is not sufficient to disable the
>>  irqfd.  The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
>>  and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
>>  
>> +On ARM/ARM64, the gsi field in the kvm_irqfd struct specifies the Shared
>> +Peripheral Interrupt (SPI) index, such that the GIC interrupt ID is
>> +given by gsi + 32.
>> +
>>  4.76 KVM_PPC_ALLOCATE_HTAB
>>  
>>  Capability: KVM_CAP_PPC_ALLOC_HTAB
>> diff --git a/arch/arm/include/uapi/asm/kvm.h b/arch/arm/include/uapi/asm/kvm.h
>> index 09ee408..77547bb 100644
>> --- a/arch/arm/include/uapi/asm/kvm.h
>> +++ b/arch/arm/include/uapi/asm/kvm.h
>> @@ -196,6 +196,9 @@ struct kvm_arch_memory_slot {
>>  /* Highest supported SPI, from VGIC_NR_IRQS */
>>  #define KVM_ARM_IRQ_GIC_MAX		127
>>  
>> +/* One single KVM irqchip, ie. the VGIC */
>> +#define KVM_NR_IRQCHIPS          1
>> +
>>  /* PSCI interface */
>>  #define KVM_PSCI_FN_BASE		0x95c1ba5e
>>  #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
>> diff --git a/arch/arm/kvm/Kconfig b/arch/arm/kvm/Kconfig
>> index 9f581b1..e519a40 100644
>> --- a/arch/arm/kvm/Kconfig
>> +++ b/arch/arm/kvm/Kconfig
>> @@ -24,6 +24,7 @@ config KVM
>>  	select KVM_MMIO
>>  	select KVM_ARM_HOST
>>  	depends on ARM_VIRT_EXT && ARM_LPAE
>> +	select HAVE_KVM_EVENTFD
>>  	---help---
>>  	  Support hosting virtualized guest machines. You will also
>>  	  need to select one or more of the processor modules below.
>> @@ -55,6 +56,7 @@ config KVM_ARM_MAX_VCPUS
>>  config KVM_ARM_VGIC
>>  	bool "KVM support for Virtual GIC"
>>  	depends on KVM_ARM_HOST && OF
>> +	select HAVE_KVM_IRQFD
>>  	default y
>>  	---help---
>>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
>> diff --git a/arch/arm/kvm/Makefile b/arch/arm/kvm/Makefile
>> index f7057ed..859db09 100644
>> --- a/arch/arm/kvm/Makefile
>> +++ b/arch/arm/kvm/Makefile
>> @@ -15,7 +15,7 @@ AFLAGS_init.o := -Wa,-march=armv7-a$(plus_virt)
>>  AFLAGS_interrupts.o := -Wa,-march=armv7-a$(plus_virt)
>>  
>>  KVM := ../../../virt/kvm
>> -kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
>> +kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
>>  
>>  obj-y += kvm-arm.o init.o interrupts.o
>>  obj-y += arm.o handle_exit.o guest.o mmu.o emulate.o reset.o
>> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
>> index 9c905b4..8f8c1da 100644
>> --- a/arch/arm/kvm/arm.c
>> +++ b/arch/arm/kvm/arm.c
>> @@ -172,6 +172,11 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>>  	case KVM_CAP_IRQCHIP:
>>  		r = vgic_present;
>>  		break;
>> +#ifdef CONFIG_HAVE_KVM_IRQFD
>> +	case KVM_CAP_IRQFD:
>> +		r = vgic_present;
>> +		break;
>> +#endif
>>  	case KVM_CAP_DEVICE_CTRL:
>>  	case KVM_CAP_USER_MEMORY:
>>  	case KVM_CAP_SYNC_MMU:
>> diff --git a/arch/arm64/include/uapi/asm/kvm.h b/arch/arm64/include/uapi/asm/kvm.h
>> index 8e38878..1ed4417 100644
>> --- a/arch/arm64/include/uapi/asm/kvm.h
>> +++ b/arch/arm64/include/uapi/asm/kvm.h
>> @@ -182,6 +182,9 @@ struct kvm_arch_memory_slot {
>>  /* Highest supported SPI, from VGIC_NR_IRQS */
>>  #define KVM_ARM_IRQ_GIC_MAX		127
>>  
>> +/* One single KVM irqchip, ie. the VGIC */
>> +#define KVM_NR_IRQCHIPS          1
>> +
>>  /* PSCI interface */
>>  #define KVM_PSCI_FN_BASE		0x95c1ba5e
>>  #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
>> diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
>> index 279e1a0..09c25c2 100644
>> --- a/arch/arm64/kvm/Kconfig
>> +++ b/arch/arm64/kvm/Kconfig
>> @@ -26,6 +26,7 @@ config KVM
>>  	select KVM_ARM_HOST
>>  	select KVM_ARM_VGIC
>>  	select KVM_ARM_TIMER
>> +	select HAVE_KVM_EVENTFD
>>  	---help---
>>  	  Support hosting virtualized guest machines.
>>  
>> @@ -50,6 +51,7 @@ config KVM_ARM_MAX_VCPUS
>>  config KVM_ARM_VGIC
>>  	bool
>>  	depends on KVM_ARM_HOST && OF
>> +	select HAVE_KVM_IRQFD
>>  	---help---
>>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
>>  
>> diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
>> index 32a0961..2e6b827 100644
>> --- a/arch/arm64/kvm/Makefile
>> +++ b/arch/arm64/kvm/Makefile
>> @@ -11,7 +11,7 @@ ARM=../../../arch/arm/kvm
>>  
>>  obj-$(CONFIG_KVM_ARM_HOST) += kvm.o
>>  
>> -kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
>> +kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
>>  kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/arm.o $(ARM)/mmu.o $(ARM)/mmio.o
>>  kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/psci.o $(ARM)/perf.o
>>  
>> diff --git a/virt/kvm/arm/vgic.c b/virt/kvm/arm/vgic.c
>> index 03affc7..d1a5c70 100644
>> --- a/virt/kvm/arm/vgic.c
>> +++ b/virt/kvm/arm/vgic.c
>> @@ -1447,7 +1447,10 @@ epilog:
>>  static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>  {
>>  	u32 status = vgic_get_interrupt_status(vcpu);
>> +	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
>>  	bool level_pending = false;
>> +	struct kvm *kvm = vcpu->kvm;
>> +	bool has_notifier;
>>  
>>  	kvm_debug("STATUS = %08x\n", status);
>>  
>> @@ -1464,6 +1467,7 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>  			struct vgic_lr vlr = vgic_get_lr(vcpu, lr);
>>  			WARN_ON(vgic_irq_is_edge(vcpu, vlr.irq));
>>  
>> +			spin_lock(&dist->lock);
>>  			vgic_irq_clear_queued(vcpu, vlr.irq);
>>  			WARN_ON(vlr.state & LR_STATE_MASK);
>>  			vlr.state = 0;
>> @@ -1482,6 +1486,24 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>  			 */
>>  			vgic_dist_irq_clear_soft_pend(vcpu, vlr.irq);
>>  
>> +			/*
>> +			 * kvm_notify_acked_irq calls kvm_set_irq()
>> +			 * to reset the IRQ level. Need to release the
>> +			 * lock for kvm_set_irq to grab it.
>> +			 */
>> +			spin_unlock(&dist->lock);
>> +
>> +			has_notifier = kvm_irq_has_notifier(kvm, 0,
>> +						vlr.irq - VGIC_NR_PRIVATE_IRQS);
>> +
>> +			if (has_notifier) {
> 
> Could that be racy? Can the irqfd be de-assigned meanwhile? Or am I
> seeing ghosts here? ;-)
Well actually kvm_irq_has_notifier implements the same loop as
kvm_notify_acked_irq  but instead of calling the irq_acked() function of
the notifier it returns true. So the check is done twice. I think I can
simply drop the kvm_irq_has_notifier check.

> 
>> +				kvm_debug("Guest EOIed vIRQ %d on CPU %d\n",
>> +					  vlr.irq, vcpu->vcpu_id);
>> +				kvm_notify_acked_irq(kvm, 0,
>> +						vlr.irq - VGIC_NR_PRIVATE_IRQS);
>> +			}
>> +			spin_lock(&dist->lock);
>> +
>>  			/* Any additional pending interrupt? */
>>  			if (vgic_dist_irq_get_level(vcpu, vlr.irq)) {
>>  				vgic_cpu_irq_set(vcpu, vlr.irq);
>> @@ -1491,6 +1513,8 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>  				vgic_cpu_irq_clear(vcpu, vlr.irq);
>>  			}
>>  
>> +			spin_unlock(&dist->lock);
>> +
>>  			/*
>>  			 * Despite being EOIed, the LR may not have
>>  			 * been marked as empty.
>> @@ -1555,14 +1579,10 @@ void kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
>>  
>>  void kvm_vgic_sync_hwstate(struct kvm_vcpu *vcpu)
>>  {
>> -	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
>> -
>>  	if (!irqchip_in_kernel(vcpu->kvm))
>>  		return;
>>  
>> -	spin_lock(&dist->lock);
>>  	__kvm_vgic_sync_hwstate(vcpu);
>> -	spin_unlock(&dist->lock);
>>  }
> 
> Given that you drop the coarse grained locking around
> __kvm_vgic_sync_hwstate(), can you adjust the comment on top of it and
> either delete it or relax the locking requirement?
yes I will remove the comment.
> 
> Have I got this right that most of the locking is not needed as long as
> we deal only with VCPU specific data structures, as any preemption would
> not execute code handling this very VCPU again?
yes this is my understanding
> 
> 
> Also it would be beneficial to note the change of the locking in the
> commit message (or even make it a separate patch). That would help
> possible bisecting.
I will add in the commit message since it looks difficult to me to
separate the lock changes from the kvm_notify_acked_irq call addition
> 
>>  
>>  int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)
>> @@ -2481,3 +2501,38 @@ out_free_irq:
>>  	free_percpu_irq(vgic->maint_irq, kvm_get_running_vcpus());
>>  	return ret;
>>  }
>> +
>> +int kvm_irq_map_gsi(struct kvm *kvm,
>> +		    struct kvm_kernel_irq_routing_entry *entries,
>> +		    int gsi)
>> +{
>> +	return gsi;
>> +}
>> +
>> +int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
>> +{
>> +	return pin;
>> +}
>> +
>> +int kvm_set_irq(struct kvm *kvm, int irq_source_id,
>> +		u32 irq, int level, bool line_status)
>> +{
>> +	unsigned int spi = irq + VGIC_NR_PRIVATE_IRQS;
>> +
>> +	kvm_debug("irqfd sets vIRQ %d to %d\n", irq, level);
>> +
>> +	if (likely(vgic_initialized(kvm))) {
> 
> Is that really needed? I reckon we can only get here if an irqfd has
> been assigned, which is guarded already by that initialization check.
Effectively we can remove it.
> 
>> +		if (spi > kvm->arch.vgic.nr_irqs)
> 
> I wonder if that check could be done earlier at assignment time. Does
> that not work because the assignment is generic code and has no notion
> of the maximum supported IRQ numbers as the VGIC has?
yes that's the reason.

Many thanks for the review of the whole series.

Best Regards

Eric
> 
> Cheers,
> Andre.
> 
>> +			return -EINVAL;
>> +		return kvm_vgic_inject_irq(kvm, 0, spi, level);
>> +	} else
>> +		return -ENODEV;
>> +}
>> +
>> +/* MSI not implemented yet */
>> +int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e,
>> +		struct kvm *kvm, int irq_source_id,
>> +		int level, bool line_status)
>> +{
>> +	return 0;
>> +}
>>
> 
> --
> To unsubscribe from this list: send the line "unsubscribe kvm" in
> the body of a message to majordomo at vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

[PATCH v7 4/4] KVM: arm/arm64: add irqfd support

[Eric Auger]

On 01/19/2015 04:01 PM, Eric Auger wrote:
> Hi Andre,
> On 01/16/2015 11:07 AM, Andr? Przywara wrote:
>> Hi Eric,
>>
>> On 01/15/2015 02:47 PM, Eric Auger wrote:
>>> This patch enables irqfd on arm/arm64.
>>>
>>> Both irqfd and resamplefd are supported. Injection is implemented
>>> in vgic.c without routing.
>>>
>>> This patch enables CONFIG_HAVE_KVM_EVENTFD and CONFIG_HAVE_KVM_IRQFD.
>>>
>>> KVM_CAP_IRQFD is now advertised. KVM_CAP_IRQFD_RESAMPLE capability
>>> automatically is advertised as soon as CONFIG_HAVE_KVM_IRQFD is set.
>>>
>>> Irqfd injection is restricted to SPI. The rationale behind not
>>> supporting PPI irqfd injection is that any device using a PPI would
>>> be a private-to-the-CPU device (timer for instance), so its state
>>> would have to be context-switched along with the VCPU and would
>>> require in-kernel wiring anyhow. It is not a relevant use case for
>>> irqfds.
>>>
>>> Signed-off-by: Eric Auger <eric.auger@linaro.org>
>>> Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
>>>
>>> ---
>>>
>>> v5 -> v6:
>>> - KVM_CAP_IRQFD support depends on vgic_present
>>> - add Christoffer's Reviewed-by
>>>
>>> v4 -> v5:
>>> - squash [PATCH v4 3/3] KVM: arm64: add irqfd support into this patch
>>> - some rewording in Documentation/virtual/kvm/api.txt and in vgic
>>>   vgic_process_maintenance unlock comment.
>>> - move explanation of why not supporting PPI into commit message
>>> - in case of injection before gic readiness, -ENODEV is returned. It is
>>>   up to the user space to avoid this situation.
>>>
>>> v3 -> v4:
>>> - reword commit message
>>> - explain why we unlock the distributor before calling kvm_notify_acked_irq
>>> - rename is_assigned_irq into has_notifier
>>> - change EOI and injection kvm_debug format string
>>> - remove error local variable in kvm_set_irq
>>> - Move HAVE_KVM_IRQCHIP unset in a separate patch
>>> - handle case were the irqfd injection is attempted before the vgic is ready.
>>>   in such a case the notifier, if any, is called immediatly
>>> - use nr_irqs to test spi is within correct range
>>>
>>> v2 -> v3:
>>> - removal of irq.h from eventfd.c put in a separate patch to increase
>>>   visibility
>>> - properly expose KVM_CAP_IRQFD capability in arm.c
>>> - remove CONFIG_HAVE_KVM_IRQCHIP meaningfull only if irq_comm.c is used
>>>
>>> v1 -> v2:
>>> - rebase on 3.17rc1
>>> - move of the dist unlock in process_maintenance
>>> - remove of dist lock in __kvm_vgic_sync_hwstate
>>> - rewording of the commit message (add resamplefd reference)
>>> - remove irq.h
>>> ---
>>>  Documentation/virtual/kvm/api.txt |  6 +++-
>>>  arch/arm/include/uapi/asm/kvm.h   |  3 ++
>>>  arch/arm/kvm/Kconfig              |  2 ++
>>>  arch/arm/kvm/Makefile             |  2 +-
>>>  arch/arm/kvm/arm.c                |  5 ++++
>>>  arch/arm64/include/uapi/asm/kvm.h |  3 ++
>>>  arch/arm64/kvm/Kconfig            |  2 ++
>>>  arch/arm64/kvm/Makefile           |  2 +-
>>>  virt/kvm/arm/vgic.c               | 63 ++++++++++++++++++++++++++++++++++++---
>>>  9 files changed, 81 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
>>> index 0007fef..5ed8088 100644
>>> --- a/Documentation/virtual/kvm/api.txt
>>> +++ b/Documentation/virtual/kvm/api.txt
>>> @@ -2231,7 +2231,7 @@ into the hash PTE second double word).
>>>  4.75 KVM_IRQFD
>>>  
>>>  Capability: KVM_CAP_IRQFD
>>> -Architectures: x86 s390
>>> +Architectures: x86 s390 arm arm64
>>>  Type: vm ioctl
>>>  Parameters: struct kvm_irqfd (in)
>>>  Returns: 0 on success, -1 on error
>>> @@ -2257,6 +2257,10 @@ Note that closing the resamplefd is not sufficient to disable the
>>>  irqfd.  The KVM_IRQFD_FLAG_RESAMPLE is only necessary on assignment
>>>  and need not be specified with KVM_IRQFD_FLAG_DEASSIGN.
>>>  
>>> +On ARM/ARM64, the gsi field in the kvm_irqfd struct specifies the Shared
>>> +Peripheral Interrupt (SPI) index, such that the GIC interrupt ID is
>>> +given by gsi + 32.
>>> +
>>>  4.76 KVM_PPC_ALLOCATE_HTAB
>>>  
>>>  Capability: KVM_CAP_PPC_ALLOC_HTAB
>>> diff --git a/arch/arm/include/uapi/asm/kvm.h b/arch/arm/include/uapi/asm/kvm.h
>>> index 09ee408..77547bb 100644
>>> --- a/arch/arm/include/uapi/asm/kvm.h
>>> +++ b/arch/arm/include/uapi/asm/kvm.h
>>> @@ -196,6 +196,9 @@ struct kvm_arch_memory_slot {
>>>  /* Highest supported SPI, from VGIC_NR_IRQS */
>>>  #define KVM_ARM_IRQ_GIC_MAX		127
>>>  
>>> +/* One single KVM irqchip, ie. the VGIC */
>>> +#define KVM_NR_IRQCHIPS          1
>>> +
>>>  /* PSCI interface */
>>>  #define KVM_PSCI_FN_BASE		0x95c1ba5e
>>>  #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
>>> diff --git a/arch/arm/kvm/Kconfig b/arch/arm/kvm/Kconfig
>>> index 9f581b1..e519a40 100644
>>> --- a/arch/arm/kvm/Kconfig
>>> +++ b/arch/arm/kvm/Kconfig
>>> @@ -24,6 +24,7 @@ config KVM
>>>  	select KVM_MMIO
>>>  	select KVM_ARM_HOST
>>>  	depends on ARM_VIRT_EXT && ARM_LPAE
>>> +	select HAVE_KVM_EVENTFD
>>>  	---help---
>>>  	  Support hosting virtualized guest machines. You will also
>>>  	  need to select one or more of the processor modules below.
>>> @@ -55,6 +56,7 @@ config KVM_ARM_MAX_VCPUS
>>>  config KVM_ARM_VGIC
>>>  	bool "KVM support for Virtual GIC"
>>>  	depends on KVM_ARM_HOST && OF
>>> +	select HAVE_KVM_IRQFD
>>>  	default y
>>>  	---help---
>>>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
>>> diff --git a/arch/arm/kvm/Makefile b/arch/arm/kvm/Makefile
>>> index f7057ed..859db09 100644
>>> --- a/arch/arm/kvm/Makefile
>>> +++ b/arch/arm/kvm/Makefile
>>> @@ -15,7 +15,7 @@ AFLAGS_init.o := -Wa,-march=armv7-a$(plus_virt)
>>>  AFLAGS_interrupts.o := -Wa,-march=armv7-a$(plus_virt)
>>>  
>>>  KVM := ../../../virt/kvm
>>> -kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
>>> +kvm-arm-y = $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
>>>  
>>>  obj-y += kvm-arm.o init.o interrupts.o
>>>  obj-y += arm.o handle_exit.o guest.o mmu.o emulate.o reset.o
>>> diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
>>> index 9c905b4..8f8c1da 100644
>>> --- a/arch/arm/kvm/arm.c
>>> +++ b/arch/arm/kvm/arm.c
>>> @@ -172,6 +172,11 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
>>>  	case KVM_CAP_IRQCHIP:
>>>  		r = vgic_present;
>>>  		break;
>>> +#ifdef CONFIG_HAVE_KVM_IRQFD
>>> +	case KVM_CAP_IRQFD:
>>> +		r = vgic_present;
>>> +		break;
>>> +#endif
>>>  	case KVM_CAP_DEVICE_CTRL:
>>>  	case KVM_CAP_USER_MEMORY:
>>>  	case KVM_CAP_SYNC_MMU:
>>> diff --git a/arch/arm64/include/uapi/asm/kvm.h b/arch/arm64/include/uapi/asm/kvm.h
>>> index 8e38878..1ed4417 100644
>>> --- a/arch/arm64/include/uapi/asm/kvm.h
>>> +++ b/arch/arm64/include/uapi/asm/kvm.h
>>> @@ -182,6 +182,9 @@ struct kvm_arch_memory_slot {
>>>  /* Highest supported SPI, from VGIC_NR_IRQS */
>>>  #define KVM_ARM_IRQ_GIC_MAX		127
>>>  
>>> +/* One single KVM irqchip, ie. the VGIC */
>>> +#define KVM_NR_IRQCHIPS          1
>>> +
>>>  /* PSCI interface */
>>>  #define KVM_PSCI_FN_BASE		0x95c1ba5e
>>>  #define KVM_PSCI_FN(n)			(KVM_PSCI_FN_BASE + (n))
>>> diff --git a/arch/arm64/kvm/Kconfig b/arch/arm64/kvm/Kconfig
>>> index 279e1a0..09c25c2 100644
>>> --- a/arch/arm64/kvm/Kconfig
>>> +++ b/arch/arm64/kvm/Kconfig
>>> @@ -26,6 +26,7 @@ config KVM
>>>  	select KVM_ARM_HOST
>>>  	select KVM_ARM_VGIC
>>>  	select KVM_ARM_TIMER
>>> +	select HAVE_KVM_EVENTFD
>>>  	---help---
>>>  	  Support hosting virtualized guest machines.
>>>  
>>> @@ -50,6 +51,7 @@ config KVM_ARM_MAX_VCPUS
>>>  config KVM_ARM_VGIC
>>>  	bool
>>>  	depends on KVM_ARM_HOST && OF
>>> +	select HAVE_KVM_IRQFD
>>>  	---help---
>>>  	  Adds support for a hardware assisted, in-kernel GIC emulation.
>>>  
>>> diff --git a/arch/arm64/kvm/Makefile b/arch/arm64/kvm/Makefile
>>> index 32a0961..2e6b827 100644
>>> --- a/arch/arm64/kvm/Makefile
>>> +++ b/arch/arm64/kvm/Makefile
>>> @@ -11,7 +11,7 @@ ARM=../../../arch/arm/kvm
>>>  
>>>  obj-$(CONFIG_KVM_ARM_HOST) += kvm.o
>>>  
>>> -kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o
>>> +kvm-$(CONFIG_KVM_ARM_HOST) += $(KVM)/kvm_main.o $(KVM)/coalesced_mmio.o $(KVM)/eventfd.o
>>>  kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/arm.o $(ARM)/mmu.o $(ARM)/mmio.o
>>>  kvm-$(CONFIG_KVM_ARM_HOST) += $(ARM)/psci.o $(ARM)/perf.o
>>>  
>>> diff --git a/virt/kvm/arm/vgic.c b/virt/kvm/arm/vgic.c
>>> index 03affc7..d1a5c70 100644
>>> --- a/virt/kvm/arm/vgic.c
>>> +++ b/virt/kvm/arm/vgic.c
>>> @@ -1447,7 +1447,10 @@ epilog:
>>>  static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>>  {
>>>  	u32 status = vgic_get_interrupt_status(vcpu);
>>> +	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
>>>  	bool level_pending = false;
>>> +	struct kvm *kvm = vcpu->kvm;
>>> +	bool has_notifier;
>>>  
>>>  	kvm_debug("STATUS = %08x\n", status);
>>>  
>>> @@ -1464,6 +1467,7 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>>  			struct vgic_lr vlr = vgic_get_lr(vcpu, lr);
>>>  			WARN_ON(vgic_irq_is_edge(vcpu, vlr.irq));
>>>  
>>> +			spin_lock(&dist->lock);
>>>  			vgic_irq_clear_queued(vcpu, vlr.irq);
>>>  			WARN_ON(vlr.state & LR_STATE_MASK);
>>>  			vlr.state = 0;
>>> @@ -1482,6 +1486,24 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>>  			 */
>>>  			vgic_dist_irq_clear_soft_pend(vcpu, vlr.irq);
>>>  
>>> +			/*
>>> +			 * kvm_notify_acked_irq calls kvm_set_irq()
>>> +			 * to reset the IRQ level. Need to release the
>>> +			 * lock for kvm_set_irq to grab it.
>>> +			 */
>>> +			spin_unlock(&dist->lock);
>>> +
>>> +			has_notifier = kvm_irq_has_notifier(kvm, 0,
>>> +						vlr.irq - VGIC_NR_PRIVATE_IRQS);
>>> +
>>> +			if (has_notifier) {
>>
>> Could that be racy? Can the irqfd be de-assigned meanwhile? Or am I
>> seeing ghosts here? ;-)
> Well actually kvm_irq_has_notifier implements the same loop as
> kvm_notify_acked_irq  but instead of calling the irq_acked() function of
> the notifier it returns true. So the check is done twice. I think I can
> simply drop the kvm_irq_has_notifier check.
> 
>>
>>> +				kvm_debug("Guest EOIed vIRQ %d on CPU %d\n",
>>> +					  vlr.irq, vcpu->vcpu_id);
>>> +				kvm_notify_acked_irq(kvm, 0,
>>> +						vlr.irq - VGIC_NR_PRIVATE_IRQS);
>>> +			}
>>> +			spin_lock(&dist->lock);
>>> +
>>>  			/* Any additional pending interrupt? */
>>>  			if (vgic_dist_irq_get_level(vcpu, vlr.irq)) {
>>>  				vgic_cpu_irq_set(vcpu, vlr.irq);
>>> @@ -1491,6 +1513,8 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)
>>>  				vgic_cpu_irq_clear(vcpu, vlr.irq);
>>>  			}
>>>  
>>> +			spin_unlock(&dist->lock);
>>> +
>>>  			/*
>>>  			 * Despite being EOIed, the LR may not have
>>>  			 * been marked as empty.
>>> @@ -1555,14 +1579,10 @@ void kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
>>>  
>>>  void kvm_vgic_sync_hwstate(struct kvm_vcpu *vcpu)
>>>  {
>>> -	struct vgic_dist *dist = &vcpu->kvm->arch.vgic;
>>> -
>>>  	if (!irqchip_in_kernel(vcpu->kvm))
>>>  		return;
>>>  
>>> -	spin_lock(&dist->lock);
>>>  	__kvm_vgic_sync_hwstate(vcpu);
>>> -	spin_unlock(&dist->lock);
>>>  }
>>
>> Given that you drop the coarse grained locking around
>> __kvm_vgic_sync_hwstate(), can you adjust the comment on top of it and
>> either delete it or relax the locking requirement?
> yes I will remove the comment.
>>
>> Have I got this right that most of the locking is not needed as long as
>> we deal only with VCPU specific data structures, as any preemption would
>> not execute code handling this very VCPU again?
> yes this is my understanding
>>
>>
>> Also it would be beneficial to note the change of the locking in the
>> commit message (or even make it a separate patch). That would help
>> possible bisecting.
> I will add in the commit message since it looks difficult to me to
> separate the lock changes from the kvm_notify_acked_irq call addition
well actually I will add a new commit ;-)

Eric
>>
>>>  
>>>  int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)
>>> @@ -2481,3 +2501,38 @@ out_free_irq:
>>>  	free_percpu_irq(vgic->maint_irq, kvm_get_running_vcpus());
>>>  	return ret;
>>>  }
>>> +
>>> +int kvm_irq_map_gsi(struct kvm *kvm,
>>> +		    struct kvm_kernel_irq_routing_entry *entries,
>>> +		    int gsi)
>>> +{
>>> +	return gsi;
>>> +}
>>> +
>>> +int kvm_irq_map_chip_pin(struct kvm *kvm, unsigned irqchip, unsigned pin)
>>> +{
>>> +	return pin;
>>> +}
>>> +
>>> +int kvm_set_irq(struct kvm *kvm, int irq_source_id,
>>> +		u32 irq, int level, bool line_status)
>>> +{
>>> +	unsigned int spi = irq + VGIC_NR_PRIVATE_IRQS;
>>> +
>>> +	kvm_debug("irqfd sets vIRQ %d to %d\n", irq, level);
>>> +
>>> +	if (likely(vgic_initialized(kvm))) {
>>
>> Is that really needed? I reckon we can only get here if an irqfd has
>> been assigned, which is guarded already by that initialization check.
> Effectively we can remove it.
>>
>>> +		if (spi > kvm->arch.vgic.nr_irqs)
>>
>> I wonder if that check could be done earlier at assignment time. Does
>> that not work because the assignment is generic code and has no notion
>> of the maximum supported IRQ numbers as the VGIC has?
> yes that's the reason.
> 
> Many thanks for the review of the whole series.
> 
> Best Regards
> 
> Eric
>>
>> Cheers,
>> Andre.
>>
>>> +			return -EINVAL;
>>> +		return kvm_vgic_inject_irq(kvm, 0, spi, level);
>>> +	} else
>>> +		return -ENODEV;
>>> +}
>>> +
>>> +/* MSI not implemented yet */
>>> +int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e,
>>> +		struct kvm *kvm, int irq_source_id,
>>> +		int level, bool line_status)
>>> +{
>>> +	return 0;
>>> +}
>>>
>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe kvm" in
>> the body of a message to majordomo at vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
>