Commit c519265f authored by Radim Krčmář's avatar Radim Krčmář Committed by Paolo Bonzini

KVM: x86: add a flag to disable KVM x2apic broadcast quirk

Add KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK as a feature flag to
KVM_CAP_X2APIC_API.

The quirk made KVM interpret 0xff as a broadcast even in x2APIC mode.
The enableable capability is needed in order to support standard x2APIC and
remain backward compatible.
Signed-off-by: default avatarRadim Krčmář <rkrcmar@redhat.com>
[Expand kvm_apic_mda comment. - Paolo]
Signed-off-by: default avatarPaolo Bonzini <pbonzini@redhat.com>
parent 37131313
...@@ -3844,12 +3844,18 @@ Returns: 0 on success, -EINVAL when args[0] contains invalid features ...@@ -3844,12 +3844,18 @@ Returns: 0 on success, -EINVAL when args[0] contains invalid features
Valid feature flags in args[0] are Valid feature flags in args[0] are
#define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0) #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0)
#define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1)
Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of Enabling KVM_X2APIC_API_USE_32BIT_IDS changes the behavior of
KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC, KVM_SET_GSI_ROUTING, KVM_SIGNAL_MSI, KVM_SET_LAPIC, and KVM_GET_LAPIC,
allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their allowing the use of 32-bit APIC IDs. See KVM_CAP_X2APIC_API in their
respective sections. respective sections.
KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK must be enabled for x2APIC to work
in logical mode or with more than 255 VCPUs. Otherwise, KVM treats 0xff
as a broadcast even in x2APIC mode in order to support physical x2APIC
without interrupt remapping. This is undesirable in logical mode,
where 0xff represents CPUs 0-7 in cluster 0.
8. Other capabilities. 8. Other capabilities.
......
...@@ -784,6 +784,7 @@ struct kvm_arch { ...@@ -784,6 +784,7 @@ struct kvm_arch {
struct page *avic_physical_id_table_page; struct page *avic_physical_id_table_page;
bool x2apic_format; bool x2apic_format;
bool x2apic_broadcast_quirk_disabled;
}; };
struct kvm_vm_stat { struct kvm_vm_stat {
......
...@@ -616,17 +616,30 @@ static bool kvm_apic_match_logical_addr(struct kvm_lapic *apic, u32 mda) ...@@ -616,17 +616,30 @@ static bool kvm_apic_match_logical_addr(struct kvm_lapic *apic, u32 mda)
} }
} }
/* KVM APIC implementation has two quirks /* The KVM local APIC implementation has two quirks:
* - dest always begins at 0 while xAPIC MDA has offset 24, *
* - IOxAPIC messages have to be delivered (directly) to x2APIC. * - the xAPIC MDA stores the destination at bits 24-31, while this
* is not true of struct kvm_lapic_irq's dest_id field. This is
* just a quirk in the API and is not problematic.
*
* - in-kernel IOAPIC messages have to be delivered directly to
* x2APIC, because the kernel does not support interrupt remapping.
* In order to support broadcast without interrupt remapping, x2APIC
* rewrites the destination of non-IPI messages from APIC_BROADCAST
* to X2APIC_BROADCAST.
*
* The broadcast quirk can be disabled with KVM_CAP_X2APIC_API. This is
* important when userspace wants to use x2APIC-format MSIs, because
* APIC_BROADCAST (0xff) is a legal route for "cluster 0, CPUs 0-7".
*/ */
static u32 kvm_apic_mda(unsigned int dest_id, struct kvm_lapic *source, static u32 kvm_apic_mda(struct kvm_vcpu *vcpu, unsigned int dest_id,
struct kvm_lapic *target) struct kvm_lapic *source, struct kvm_lapic *target)
{ {
bool ipi = source != NULL; bool ipi = source != NULL;
bool x2apic_mda = apic_x2apic_mode(ipi ? source : target); bool x2apic_mda = apic_x2apic_mode(ipi ? source : target);
if (!ipi && dest_id == APIC_BROADCAST && x2apic_mda) if (!vcpu->kvm->arch.x2apic_broadcast_quirk_disabled &&
!ipi && dest_id == APIC_BROADCAST && x2apic_mda)
return X2APIC_BROADCAST; return X2APIC_BROADCAST;
return x2apic_mda ? dest_id : SET_APIC_DEST_FIELD(dest_id); return x2apic_mda ? dest_id : SET_APIC_DEST_FIELD(dest_id);
...@@ -636,7 +649,7 @@ bool kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source, ...@@ -636,7 +649,7 @@ bool kvm_apic_match_dest(struct kvm_vcpu *vcpu, struct kvm_lapic *source,
int short_hand, unsigned int dest, int dest_mode) int short_hand, unsigned int dest, int dest_mode)
{ {
struct kvm_lapic *target = vcpu->arch.apic; struct kvm_lapic *target = vcpu->arch.apic;
u32 mda = kvm_apic_mda(dest, source, target); u32 mda = kvm_apic_mda(vcpu, dest, source, target);
apic_debug("target %p, source %p, dest 0x%x, " apic_debug("target %p, source %p, dest 0x%x, "
"dest_mode 0x%x, short_hand 0x%x\n", "dest_mode 0x%x, short_hand 0x%x\n",
...@@ -688,6 +701,25 @@ static void kvm_apic_disabled_lapic_found(struct kvm *kvm) ...@@ -688,6 +701,25 @@ static void kvm_apic_disabled_lapic_found(struct kvm *kvm)
} }
} }
static bool kvm_apic_is_broadcast_dest(struct kvm *kvm, struct kvm_lapic **src,
struct kvm_lapic_irq *irq, struct kvm_apic_map *map)
{
if (kvm->arch.x2apic_broadcast_quirk_disabled) {
if ((irq->dest_id == APIC_BROADCAST &&
map->mode != KVM_APIC_MODE_X2APIC))
return true;
if (irq->dest_id == X2APIC_BROADCAST)
return true;
} else {
bool x2apic_ipi = src && *src && apic_x2apic_mode(*src);
if (irq->dest_id == (x2apic_ipi ?
X2APIC_BROADCAST : APIC_BROADCAST))
return true;
}
return false;
}
/* Return true if the interrupt can be handled by using *bitmap as index mask /* Return true if the interrupt can be handled by using *bitmap as index mask
* for valid destinations in *dst array. * for valid destinations in *dst array.
* Return false if kvm_apic_map_get_dest_lapic did nothing useful. * Return false if kvm_apic_map_get_dest_lapic did nothing useful.
...@@ -701,7 +733,6 @@ static inline bool kvm_apic_map_get_dest_lapic(struct kvm *kvm, ...@@ -701,7 +733,6 @@ static inline bool kvm_apic_map_get_dest_lapic(struct kvm *kvm,
unsigned long *bitmap) unsigned long *bitmap)
{ {
int i, lowest; int i, lowest;
bool x2apic_ipi;
if (irq->shorthand == APIC_DEST_SELF && src) { if (irq->shorthand == APIC_DEST_SELF && src) {
*dst = src; *dst = src;
...@@ -710,11 +741,7 @@ static inline bool kvm_apic_map_get_dest_lapic(struct kvm *kvm, ...@@ -710,11 +741,7 @@ static inline bool kvm_apic_map_get_dest_lapic(struct kvm *kvm,
} else if (irq->shorthand) } else if (irq->shorthand)
return false; return false;
x2apic_ipi = src && *src && apic_x2apic_mode(*src); if (!map || kvm_apic_is_broadcast_dest(kvm, src, irq, map))
if (irq->dest_id == (x2apic_ipi ? X2APIC_BROADCAST : APIC_BROADCAST))
return false;
if (!map)
return false; return false;
if (irq->dest_mode == APIC_DEST_PHYSICAL) { if (irq->dest_mode == APIC_DEST_PHYSICAL) {
......
...@@ -90,7 +90,8 @@ static u64 __read_mostly efer_reserved_bits = ~((u64)EFER_SCE); ...@@ -90,7 +90,8 @@ static u64 __read_mostly efer_reserved_bits = ~((u64)EFER_SCE);
#define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM #define VM_STAT(x) offsetof(struct kvm, stat.x), KVM_STAT_VM
#define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU #define VCPU_STAT(x) offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU
#define KVM_X2APIC_API_VALID_FLAGS (KVM_X2APIC_API_USE_32BIT_IDS) #define KVM_X2APIC_API_VALID_FLAGS (KVM_X2APIC_API_USE_32BIT_IDS | \
KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK)
static void update_cr8_intercept(struct kvm_vcpu *vcpu); static void update_cr8_intercept(struct kvm_vcpu *vcpu);
static void process_nmi(struct kvm_vcpu *vcpu); static void process_nmi(struct kvm_vcpu *vcpu);
...@@ -3811,6 +3812,8 @@ static int kvm_vm_ioctl_enable_cap(struct kvm *kvm, ...@@ -3811,6 +3812,8 @@ static int kvm_vm_ioctl_enable_cap(struct kvm *kvm,
if (cap->args[0] & KVM_X2APIC_API_USE_32BIT_IDS) if (cap->args[0] & KVM_X2APIC_API_USE_32BIT_IDS)
kvm->arch.x2apic_format = true; kvm->arch.x2apic_format = true;
if (cap->args[0] & KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK)
kvm->arch.x2apic_broadcast_quirk_disabled = true;
r = 0; r = 0;
break; break;
......
...@@ -1315,5 +1315,6 @@ struct kvm_assigned_msix_entry { ...@@ -1315,5 +1315,6 @@ struct kvm_assigned_msix_entry {
}; };
#define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0) #define KVM_X2APIC_API_USE_32BIT_IDS (1ULL << 0)
#define KVM_X2APIC_API_DISABLE_BROADCAST_QUIRK (1ULL << 1)
#endif /* __LINUX_KVM_H */ #endif /* __LINUX_KVM_H */
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment