[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 5/6] kvm: Don't assume irqchip implies MSI routi
From: |
Jan Kiszka |
Subject: |
Re: [Qemu-devel] [PATCH 5/6] kvm: Don't assume irqchip implies MSI routing via irqfds |
Date: |
Wed, 25 Jul 2012 17:49:29 +0200 |
User-agent: |
Mozilla/5.0 (X11; U; Linux i686 (x86_64); de; rv:1.8.1.12) Gecko/20080226 SUSE/2.0.0.12-1.1 Thunderbird/2.0.0.12 Mnenhy/0.7.5.666 |
On 2012-07-25 15:24, Peter Maydell wrote:
> Decouple another x86-specific assumption about what irqchips imply.
>
> Signed-off-by: Peter Maydell <address@hidden>
> ---
> hw/virtio-pci.c | 4 ++--
> kvm-all.c | 3 ++-
> kvm-stub.c | 1 +
> kvm.h | 12 ++++++++++++
> target-i386/kvm.c | 4 +++-
> 5 files changed, 20 insertions(+), 4 deletions(-)
>
> diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
> index 4e03f0b..98e02ef 100644
> --- a/hw/virtio-pci.c
> +++ b/hw/virtio-pci.c
> @@ -627,7 +627,7 @@ static int virtio_pci_set_guest_notifiers(void *opaque,
> bool assign)
> int r, n;
>
> /* Must unset vector notifier while guest notifier is still assigned */
> - if (kvm_irqchip_in_kernel() && !assign) {
> + if (kvm_msi_via_irqfd_enabled() && !assign) {
> msix_unset_vector_notifiers(&proxy->pci_dev);
> g_free(proxy->vector_irqfd);
> proxy->vector_irqfd = NULL;
> @@ -645,7 +645,7 @@ static int virtio_pci_set_guest_notifiers(void *opaque,
> bool assign)
> }
>
> /* Must set vector notifier after guest notifier has been assigned */
> - if (kvm_irqchip_in_kernel() && assign) {
> + if (kvm_msi_via_irqfd_enabled() && assign) {
> proxy->vector_irqfd =
> g_malloc0(sizeof(*proxy->vector_irqfd) *
> msix_nr_vectors_allocated(&proxy->pci_dev));
These are ok.
> diff --git a/kvm-all.c b/kvm-all.c
> index a88b8ad..c624137 100644
> --- a/kvm-all.c
> +++ b/kvm-all.c
> @@ -102,6 +102,7 @@ KVMState *kvm_state;
> bool kvm_kernel_irqchip;
> bool kvm_async_interrupt_injection;
> bool kvm_irqfds_allowed;
> +bool kvm_msi_via_irqfd_allowed;
>
> static const KVMCapabilityInfo kvm_required_capabilites[] = {
> KVM_CAP_INFO(USER_MEMORY),
> @@ -1098,7 +1099,7 @@ int kvm_irqchip_add_msi_route(KVMState *s, MSIMessage
> msg)
> struct kvm_irq_routing_entry kroute;
> int virq;
>
> - if (!kvm_irqchip_in_kernel()) {
> + if (!kvm_msi_via_irqfd_enabled()) {
This is semantically wrong. Currently, this function is only used for
irqfd. But we will also use it to prepare MSI injections by KVM device
assignment. For other (yet non-existent) in-kernel sources of MSIs we
could use it as well. The proper check is some active version of
has_gsi_routing.
> return -ENOSYS;
> }
>
> diff --git a/kvm-stub.c b/kvm-stub.c
> index 179e5de..6cdeb1c 100644
> --- a/kvm-stub.c
> +++ b/kvm-stub.c
> @@ -21,6 +21,7 @@ KVMState *kvm_state;
> bool kvm_kernel_irqchip;
> bool kvm_async_interrupt_injection;
> bool kvm_irqfds_allowed;
> +bool kvm_msi_via_irqfd_allowed;
>
> int kvm_init_vcpu(CPUArchState *env)
> {
> diff --git a/kvm.h b/kvm.h
> index 2337eb0..1449795 100644
> --- a/kvm.h
> +++ b/kvm.h
> @@ -26,6 +26,7 @@ extern int kvm_allowed;
> extern bool kvm_kernel_irqchip;
> extern bool kvm_async_interrupt_injection;
> extern bool kvm_irqfds_allowed;
> +extern bool kvm_msi_via_irqfd_allowed;
>
> #if defined CONFIG_KVM || !defined NEED_CPU_H
> #define kvm_enabled() (kvm_allowed)
> @@ -47,11 +48,22 @@ extern bool kvm_irqfds_allowed;
> * with a configuration where it is meaningful to use them).
> */
> #define kvm_irqfds_enabled() (kvm_irqfds_allowed)
> +/**
> + * kvm_msi_via_irqfd_enabled:
> + *
> + * Returns: true if we can route a PCI MSI (Message Signaled Interrupt)
> + * to a KVM CPU via an irqfd. This requires that the kernel supports
> + * this and that we're running in a configuration that permits it.
> + * This should be checked before calling MSI related functions such as
> + * kvm_irqchip_add_msi_route.
See above.
> + */
> +#define kvm_msi_via_irqfd_enabled() (kvm_msi_via_irqfd_allowed)
> #else
> #define kvm_enabled() (0)
> #define kvm_irqchip_in_kernel() (false)
> #define kvm_async_interrupt_injection() (false)
> #define kvm_irqfds_enabled() (false)
> +#define kvm_msi_via_irqfd_enabled() (false)
> #endif
>
> struct kvm_run;
> diff --git a/target-i386/kvm.c b/target-i386/kvm.c
> index 8e19a4d..03db818 100644
> --- a/target-i386/kvm.c
> +++ b/target-i386/kvm.c
> @@ -2046,7 +2046,9 @@ void kvm_arch_init_irq_routing(KVMState *s)
> no_hpet = 1;
> }
> /* We know at this point that we're using the in-kernel
> - * irqchip, so we can use irqfds.
> + * irqchip, so we can use irqfds, and on x86 we know
> + * we can use msi via irqfd.
> */
> kvm_irqfds_allowed = true;
> + kvm_msi_via_irqfd_allowed = true;
> }
>
Jan
--
Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
Corporate Competence Center Embedded Linux
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, (continued)
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, Jan Kiszka, 2012/07/25
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, Avi Kivity, 2012/07/25
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, Peter Maydell, 2012/07/25
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, Jan Kiszka, 2012/07/25
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, Avi Kivity, 2012/07/25
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, Peter Maydell, 2012/07/25
- Re: [Qemu-devel] [PATCH 2/6] kvm: Rename kvm_irqchip_set_irq to kvm_inject_async_irq, Avi Kivity, 2012/07/25
[Qemu-devel] [PATCH 6/6] kvm: Add documentation comment for kvm_irqchip_in_kernel(), Peter Maydell, 2012/07/25
[Qemu-devel] [PATCH 5/6] kvm: Don't assume irqchip implies MSI routing via irqfds, Peter Maydell, 2012/07/25
- Re: [Qemu-devel] [PATCH 5/6] kvm: Don't assume irqchip implies MSI routing via irqfds,
Jan Kiszka <=
[Qemu-devel] [PATCH 4/6] kvm: Don't assume irqchip-in-kernel implies irqfds, Peter Maydell, 2012/07/25
[Qemu-devel] [PATCH 3/6] kvm: Move kvm_allows_irq0_override() to target-i386, Peter Maydell, 2012/07/25
[Qemu-devel] [PATCH 1/6] kvm: Decouple 'interrupt injection is async' from 'kernel irqchip', Peter Maydell, 2012/07/25