| Message ID | 20171105135233.34572-3-nikita.leshchenko@oracle.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
2017-11-05 21:52 GMT+08:00 Nikita Leshenko <nikita.leshchenko@oracle.com>: > Avoid firing a level-triggered interrupt that has the Remote IRR bit set, > because that means that some CPU is already processing it. The Remote > IRR bit will be cleared after an EOI and the interrupt will refire > if the irq line is still asserted. > > This behavior is aligned with QEMU's IOAPIC implementation that was > introduced by commit f99b86b94987 > ("x86: ioapic: ignore level irq during processing") in QEMU. > > Signed-off-by: Nikita Leshenko <nikita.leshchenko@oracle.com> > Reviewed-by: Liran Alon <liran.alon@oracle.com> > Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Reviewed-by: Wanpeng Li <wanpeng.li@hotmail.com> > --- > arch/x86/kvm/ioapic.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/arch/x86/kvm/ioapic.c b/arch/x86/kvm/ioapic.c > index ae0a7dc318b2..5c9231139243 100644 > --- a/arch/x86/kvm/ioapic.c > +++ b/arch/x86/kvm/ioapic.c > @@ -323,7 +323,9 @@ static int ioapic_service(struct kvm_ioapic *ioapic, int irq, bool line_status) > struct kvm_lapic_irq irqe; > int ret; > > - if (entry->fields.mask) > + if (entry->fields.mask || > + (entry->fields.trig_mode == IOAPIC_LEVEL_TRIG && > + entry->fields.remote_irr)) > return -1; > > ioapic_debug("dest=%x dest_mode=%x delivery_mode=%x " > -- > 2.13.3 >
diff --git a/arch/x86/kvm/ioapic.c b/arch/x86/kvm/ioapic.c index ae0a7dc318b2..5c9231139243 100644 --- a/arch/x86/kvm/ioapic.c +++ b/arch/x86/kvm/ioapic.c @@ -323,7 +323,9 @@ static int ioapic_service(struct kvm_ioapic *ioapic, int irq, bool line_status) struct kvm_lapic_irq irqe; int ret; - if (entry->fields.mask) + if (entry->fields.mask || + (entry->fields.trig_mode == IOAPIC_LEVEL_TRIG && + entry->fields.remote_irr)) return -1; ioapic_debug("dest=%x dest_mode=%x delivery_mode=%x "
Avoid firing a level-triggered interrupt that has the Remote IRR bit set, because that means that some CPU is already processing it. The Remote IRR bit will be cleared after an EOI and the interrupt will refire if the irq line is still asserted. This behavior is aligned with QEMU's IOAPIC implementation that was introduced by commit f99b86b94987 ("x86: ioapic: ignore level irq during processing") in QEMU. Signed-off-by: Nikita Leshenko <nikita.leshchenko@oracle.com> Reviewed-by: Liran Alon <liran.alon@oracle.com> Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> --- arch/x86/kvm/ioapic.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)