[PATCH-4.5 v2 09/10] xen/arm: use GICH_ELSR[01] to avoid reading all the GICH_LRs in gic_clear_lrs

[Stefano Stabellini <stefano.stabellini@eu.citrix.com>]

Read GICH_ELSR0 and GICH_ELSR1 to figure out which GICH_LR registers
do not contain valid interrupts. Only call _gic_clear_lr on those.

If a cpu is trying to inject an interrupt that is already inflight into
another cpu, it sets GIC_IRQ_GUEST_PENDING and sends an SGI to it.
The target cpu is going to be interrupted and _gic_clear_lr, called by
gic_clear_lrs, will take care of setting GICH_LR_PENDING if the irq is
active.
In order to make sure that _gic_clear_lr is called for this irq, avoid
filtering lr_mask with GICH_ELSR[01] in this case (so that in this
situation we call _gic_clear_lr on all the GICH_LRs). Use a simple
percpu bit, lr_clear_all, set by the sender cpu and reset by the
receiver cpu, to understand whether we need to evaluate all GICH_LRs or
we can filter them.

Signed-off-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
---
 xen/arch/arm/gic.c        |   27 ++++++++++++++++++++++-----
 xen/arch/arm/vgic.c       |    1 +
 xen/include/asm-arm/gic.h |    1 +
 3 files changed, 24 insertions(+), 5 deletions(-)

diff --git a/xen/arch/arm/gic.c b/xen/arch/arm/gic.c
index 54be9ca..b00f77c 100644
--- a/xen/arch/arm/gic.c
+++ b/xen/arch/arm/gic.c
@@ -55,6 +55,7 @@ static struct {
 static irq_desc_t irq_desc[NR_IRQS];
 static DEFINE_PER_CPU(irq_desc_t[NR_LOCAL_IRQS], local_irq_desc);
 static DEFINE_PER_CPU(uint64_t, lr_mask);
+static DEFINE_PER_CPU(uint8_t, lr_clear_all);
 
 static unsigned nr_lrs;
 
@@ -67,7 +68,7 @@ static DEFINE_PER_CPU(u8, gic_cpu_id);
 /* Maximum cpu interface per GIC */
 #define NR_GIC_CPU_IF 8
 
-static void gic_clear_lrs(struct vcpu *v);
+static void gic_clear_lrs(struct vcpu *v, bool_t all);
 
 static unsigned int gic_cpu_mask(const cpumask_t *cpumask)
 {
@@ -109,6 +110,7 @@ void gic_save_state(struct vcpu *v)
         v->arch.gic_lr[i] = GICH[GICH_LR + i];
     v->arch.lr_mask = this_cpu(lr_mask);
     v->arch.gic_apr = GICH[GICH_APR];
+    this_cpu(lr_clear_all) = 0ULL;
     /* Disable until next VCPU scheduled */
     GICH[GICH_HCR] = 0;
     isb();
@@ -122,13 +124,14 @@ void gic_restore_state(struct vcpu *v)
         return;
 
     this_cpu(lr_mask) = v->arch.lr_mask;
+    this_cpu(lr_clear_all) = 0ULL;
     for ( i=0; i<nr_lrs; i++)
         GICH[GICH_LR + i] = v->arch.gic_lr[i];
     GICH[GICH_APR] = v->arch.gic_apr;
     GICH[GICH_HCR] = GICH_HCR_EN;
     isb();
 
-    gic_clear_lrs(v);
+    gic_clear_lrs(v, 1);
     gic_restore_pending_irqs(v);
 }
 
@@ -372,6 +375,7 @@ static void __cpuinit gic_hyp_init(void)
 
     GICH[GICH_MISR] = GICH_MISR_EOI;
     this_cpu(lr_mask) = 0ULL;
+    this_cpu(lr_clear_all) = 0ULL;
 }
 
 static void __cpuinit gic_hyp_disable(void)
@@ -726,11 +730,19 @@ static void _gic_clear_lr(struct vcpu *v, int i, int vgic_locked)
     }
 }
 
-static void gic_clear_lrs(struct vcpu *v)
+static void gic_clear_lrs(struct vcpu *v, bool_t all)
 {
     int i = 0;
+    uint64_t elsr;
+    
+    if ( !all )
+    {
+        elsr = GICH[GICH_ELSR0] | (((uint64_t) GICH[GICH_ELSR1]) << 32);
+        elsr &= this_cpu(lr_mask);
+    } else
+        elsr = this_cpu(lr_mask);
 
-    while ((i = find_next_bit((const long unsigned int *) &this_cpu(lr_mask),
+    while ((i = find_next_bit((const long unsigned int *) &elsr,
                               nr_lrs, i)) < nr_lrs) {
 
         _gic_clear_lr(v, i, 0);
@@ -743,6 +755,11 @@ void gic_set_clear_lr(struct vcpu *v, struct pending_irq *p)
     _gic_clear_lr(v, p->lr, 1);
 }
 
+void gic_set_clear_lrs_other(struct vcpu *v)
+{
+    set_bit(0, &per_cpu(lr_clear_all, v->processor));
+}
+
 static void gic_restore_pending_irqs(struct vcpu *v)
 {
     int i;
@@ -796,7 +813,7 @@ int gic_events_need_delivery(void)
 
 void gic_inject(void)
 {
-    gic_clear_lrs(current);
+    gic_clear_lrs(current, test_and_clear_bit(0, &this_cpu(lr_clear_all)));
 
     gic_restore_pending_irqs(current);
     if (!gic_events_need_delivery())
diff --git a/xen/arch/arm/vgic.c b/xen/arch/arm/vgic.c
index 4bfab26..02ad3cd 100644
--- a/xen/arch/arm/vgic.c
+++ b/xen/arch/arm/vgic.c
@@ -716,6 +716,7 @@ void vgic_vcpu_inject_irq(struct vcpu *v, unsigned int irq)
             return;
         } else {
             set_bit(GIC_IRQ_GUEST_PENDING, &n->status);
+            gic_set_clear_lrs_other(v);
             goto out;
         }
     }
diff --git a/xen/include/asm-arm/gic.h b/xen/include/asm-arm/gic.h
index 6de0d9b..8d36f7c 100644
--- a/xen/include/asm-arm/gic.h
+++ b/xen/include/asm-arm/gic.h
@@ -185,6 +185,7 @@ extern int gic_route_irq_to_guest(struct domain *d,
                                   const struct dt_irq *irq,
                                   const char * devname);
 extern void gic_set_clear_lr(struct vcpu *v, struct pending_irq *p);
+extern void gic_set_clear_lrs_other(struct vcpu *v);
 
 /* Accept an interrupt from the GIC and dispatch its handler */
 extern void gic_interrupt(struct cpu_user_regs *regs, int is_fiq);
-- 
1.7.10.4