VGIC fixes for 3.9-rc1
cover VGIC fixes for 3.9-rc1
I've just realised that these patches have never been on LAKML, and only to kvm-arm. It is probably better to repost them here before they get pulled in.
The first patch is just a cleanup, and has very little consequence on the behaviour of the VGIC. The second one is more serious, and results in lost interrupts under heavy load.
Patches are against mainline as of today, and tested on both TC2 and arm64 model.
Cheers,
M.
Marc Zyngier (2):
ARM: KVM: vgic: force EOIed LRs to the empty state
ARM: KVM: vgic: take distributor lock on sync_hwstate path
arch/arm/kvm/vgic.c | 35 ++++++++++++++--------------------- 1 file changed, 14 insertions(+), 21 deletions(-)
Signature · 1 line
1.7.12.4
PATCH 1/2 ARM: KVM: vgic: force EOIed LRs to the empty state
The VGIC doesn't guarantee that an EOIed LR that has been configured to generate a maintenance interrupt will appear as empty.
While the code recovers from this situation, it is better to clean the LR and flag it as empty so it can be quickly recycled.
--- arch/arm/kvm/vgic.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-)
arch/arm/kvm/vgic.c +8 -3
@@ -883,8 +883,7 @@ static bool vgic_queue_irq(struct kvm_vcpu *vcpu, u8 sgi_source_id, int irq) lr, irq, vgic_cpu->vgic_lr[lr]);BUG_ON (!test_bit (lr, vgic_cpu->lr_used)); vgic_cpu->vgic_lr[lr] |=GICH_LR_PENDING_BIT ; - - goto out; +return true; }/* Try to use another LR for this interrupt */
@@ -898,7 +897,6 @@ static bool vgic_queue_irq(struct kvm_vcpu *vcpu, u8 sgi_source_id, int irq) vgic_cpu->vgic_irq_lr_map[irq] = lr;set_bit (lr, vgic_cpu->lr_used); -out:if (!vgic_irq_is_edge (vcpu, irq)) vgic_cpu->vgic_lr[lr] |=GICH_LR_EOI ;
@@ -1054,6 +1052,13 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu) }else {vgic_cpu_irq_clear (vcpu, irq); } + +/* +* Despite being EOIed, the LR may not have +* been marked as empty. +*/ +set_bit (lr, (unsigned long *)vgic_cpu->vgic_elrsr); + vgic_cpu->vgic_lr[lr] &= ~GICH_LR_ACTIVE_BIT ; } }
Signature · 1 line
1.7.12.4
PATCH 2/2 ARM: KVM: vgic: take distributor lock on sync_hwstate path
In reply to Marc ZyngierNow that the maintenance interrupt handling is actually out of the handler itself, the code becomes quite racy as we can get preempted while we process the state.
Wrapping this code around the distributor lock ensures that we're not preempted and relatively race-free.
--- arch/arm/kvm/vgic.c | 24 ++++++------------------ 1 file changed, 6 insertions(+), 18 deletions(-)
arch/arm/kvm/vgic.c +6 -18
@@ -1016,21 +1016,6 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu)kvm_debug ("MISR = %08x\n" , vgic_cpu->vgic_misr); -/* -* We do not need to take the distributor lock here, since the only -* action we perform is clearing the irq_active_bit for an EOIed -* level interrupt. There is a potential race with -* the queuing of an interrupt in __kvm_vgic_flush_hwstate(), where we -* check if the interrupt is already active. Two possibilities: -* -* - The queuing is occurring on the same vcpu: cannot happen, -* as we're already in the context of this vcpu, and -* executing the handler -* - The interrupt has been migrated to another vcpu, and we -* ignore this interrupt for this run. Big deal. It is still -* pending though, and will get considered when this vcpu -* exits. -*/ if (vgic_cpu->vgic_misr &GICH_MISR_EOI ) {/* * Some level interrupts have been EOIed. Clear their
@@ -1069,9 +1054,8 @@ static bool vgic_process_maintenance(struct kvm_vcpu *vcpu) }/* -* Sync back the VGIC state after a guest run. We do not really touch -* the distributor here (the irq_pending_on_cpu bit is safe to set), -* so there is no need for taking its lock. +* Sync back the VGIC state after a guest run. The distributor lock is +* needed so we don't get preempted in the middle of the state processing. */ static void __kvm_vgic_sync_hwstate (struct kvm_vcpu *vcpu) {
@@ -1117,10 +1101,14 @@ 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)
Signature · 1 line
1.7.12.4
cover VGIC fixes for 3.9-rc1
In reply to Marc ZyngierQuoted 10 lines · Marc Zyngier
I've just realised that these patches have never been on LAKML, and only to kvm-arm. It is probably better to repost them here before they get pulled in.
The first patch is just a cleanup, and has very little consequence on the behaviour of the VGIC. The second one is more serious, and results in lost interrupts under heavy load.
Patches are against mainline as of today, and tested on both TC2 and arm64 model.
I'm probably turning up late to this party, but after wrapping my head around the vgic once again:
for these two patches.
Will
cover VGIC fixes for 3.9-rc1
Quoted 12 lines · Will Deacon
Quoted 10 lines · Marc Zyngier
I've just realised that these patches have never been on LAKML, and only to kvm-arm. It is probably better to repost them here before they get pulled in.
The first patch is just a cleanup, and has very little consequence on the behaviour of the VGIC. The second one is more serious, and results in lost interrupts under heavy load.
Patches are against mainline as of today, and tested on both TC2 and arm64 model.
Quoted 6 lines
I'm probably turning up late to this party, but after wrapping my head around the vgic once again:
Reviewed-by Will Deacon <[email protected]>for these two patches.
Thanks Will. I have some paracetamol if you need any... ;-)
M.
Signature · 1 line
Jazz is not dead. It just smells funny...