|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [Xen-changelog] Clean up new EOI ack method some more and fix unbinding
# HG changeset patch
# User kaf24@xxxxxxxxxxxxxxxxxxxx
# Node ID 3c1cd09801c047008e529aa03b56059e00c1f4f2
# Parent 91da9a1b7196b093c6b44906992bcbcad92b30a1
Clean up new EOI ack method some more and fix unbinding
IRQ from guest (penidng EOIs must be forcibly flushed).
Signed-off-by: Keir Fraser <keir@xxxxxxxxxxxxx>
diff -r 91da9a1b7196 -r 3c1cd09801c0 xen/arch/x86/io_apic.c
--- a/xen/arch/x86/io_apic.c Sat Apr 15 18:25:21 2006
+++ b/xen/arch/x86/io_apic.c Sun Apr 16 14:04:21 2006
@@ -200,18 +200,6 @@
static void __level_IO_APIC_irq (unsigned int irq)
{
__modify_IO_APIC_irq(irq, 0x00008000, 0);
-}
-
-/* mask = 1, trigger = 0 */
-static void __mask_and_edge_IO_APIC_irq (unsigned int irq)
-{
- __modify_IO_APIC_irq(irq, 0x00010000, 0x00008000);
-}
-
-/* mask = 0, trigger = 1 */
-static void __unmask_and_level_IO_APIC_irq (unsigned int irq)
-{
- __modify_IO_APIC_irq(irq, 0x00008000, 0x00010000);
}
static void mask_IO_APIC_irq (unsigned int irq)
@@ -1395,7 +1383,8 @@
if ( !ioapic_ack_new )
{
- unmask_IO_APIC_irq(irq);
+ if ( !(irq_desc[IO_APIC_VECTOR(irq)].status & IRQ_DISABLED) )
+ unmask_IO_APIC_irq(irq);
return;
}
@@ -1427,8 +1416,11 @@
if (!(v & (1 << (i & 0x1f)))) {
atomic_inc(&irq_mis_count);
spin_lock(&ioapic_lock);
- __mask_and_edge_IO_APIC_irq(irq);
- __unmask_and_level_IO_APIC_irq(irq);
+ __mask_IO_APIC_irq(irq);
+ __edge_IO_APIC_irq(irq);
+ __level_IO_APIC_irq(irq);
+ if ( !(irq_desc[IO_APIC_VECTOR(irq)].status & IRQ_DISABLED) )
+ __unmask_IO_APIC_irq(irq);
spin_unlock(&ioapic_lock);
}
}
diff -r 91da9a1b7196 -r 3c1cd09801c0 xen/arch/x86/irq.c
--- a/xen/arch/x86/irq.c Sat Apr 15 18:25:21 2006
+++ b/xen/arch/x86/irq.c Sun Apr 16 14:04:21 2006
@@ -151,7 +151,7 @@
u8 ack_type;
#define ACKTYPE_NONE 0 /* No final acknowledgement is required */
#define ACKTYPE_UNMASK 1 /* Unmask PIC hardware (from any CPU) */
-#define ACKTYPE_LAPIC_EOI 2 /* EOI on the CPU that was interrupted */
+#define ACKTYPE_EOI 2 /* EOI on the CPU that was interrupted */
cpumask_t cpu_eoi_map; /* CPUs that need to EOI this interrupt */
struct domain *guest[IRQ_MAX_GUESTS];
} irq_guest_action_t;
@@ -161,10 +161,10 @@
* order, as only the current highest-priority pending irq can be EOIed.
*/
static struct {
- u8 vector;
- u8 ready_to_end;
-} pending_lapic_eoi[NR_CPUS][NR_VECTORS] __cacheline_aligned;
-#define pending_lapic_eoi_sp(cpu) (pending_lapic_eoi[cpu][NR_VECTORS-1].vector)
+ u8 vector; /* Vector awaiting EOI */
+ u8 ready; /* Ready for EOI now? */
+} pending_eoi[NR_CPUS][NR_VECTORS] __cacheline_aligned;
+#define pending_eoi_sp(cpu) (pending_eoi[cpu][NR_VECTORS-1].vector)
static void __do_IRQ_guest(int vector)
{
@@ -176,20 +176,21 @@
if ( unlikely(action->nr_guests == 0) )
{
- /* An interrupt may slip through while freeing a LAPIC_EOI irq. */
- ASSERT(action->ack_type == ACKTYPE_LAPIC_EOI);
+ /* An interrupt may slip through while freeing an ACKTYPE_EOI irq. */
+ ASSERT(action->ack_type == ACKTYPE_EOI);
+ ASSERT(desc->status & IRQ_DISABLED);
desc->handler->end(vector);
return;
}
- if ( action->ack_type == ACKTYPE_LAPIC_EOI )
- {
- sp = pending_lapic_eoi_sp(cpu);
- ASSERT((sp == 0) || (pending_lapic_eoi[cpu][sp-1].vector < vector));
+ if ( action->ack_type == ACKTYPE_EOI )
+ {
+ sp = pending_eoi_sp(cpu);
+ ASSERT((sp == 0) || (pending_eoi[cpu][sp-1].vector < vector));
ASSERT(sp < (NR_VECTORS-1));
- pending_lapic_eoi[cpu][sp].vector = vector;
- pending_lapic_eoi[cpu][sp].ready_to_end = 0;
- pending_lapic_eoi_sp(cpu) = sp+1;
+ pending_eoi[cpu][sp].vector = vector;
+ pending_eoi[cpu][sp].ready = 0;
+ pending_eoi_sp(cpu) = sp+1;
cpu_set(cpu, action->cpu_eoi_map);
}
@@ -203,47 +204,93 @@
}
}
-static void end_guest_irq(void *data)
-{
- irq_desc_t *desc = data;
+/* Flush all ready EOIs from the top of this CPU's pending-EOI stack. */
+static void flush_ready_eoi(void *unused)
+{
+ irq_desc_t *desc;
+ int vector, sp, cpu = smp_processor_id();
+
+ ASSERT(!local_irq_is_enabled());
+
+ sp = pending_eoi_sp(cpu);
+
+ while ( (--sp >= 0) && pending_eoi[cpu][sp].ready )
+ {
+ vector = pending_eoi[cpu][sp].vector;
+ desc = &irq_desc[vector];
+ spin_lock(&desc->lock);
+ desc->handler->end(vector);
+ spin_unlock(&desc->lock);
+ }
+
+ pending_eoi_sp(cpu) = sp+1;
+}
+
+static void __set_eoi_ready(irq_desc_t *desc)
+{
irq_guest_action_t *action = (irq_guest_action_t *)desc->action;
- unsigned long flags;
int vector, sp, cpu = smp_processor_id();
vector = desc - irq_desc;
- spin_lock_irqsave(&desc->lock, flags);
-
- if ( (desc->status & IRQ_GUEST) &&
- (action->in_flight == 0) &&
- test_and_clear_bit(cpu, &action->cpu_eoi_map) )
- {
- sp = pending_lapic_eoi_sp(cpu);
- do {
- ASSERT(sp > 0);
- } while ( pending_lapic_eoi[cpu][--sp].vector != vector );
- ASSERT(!pending_lapic_eoi[cpu][sp].ready_to_end);
- pending_lapic_eoi[cpu][sp].ready_to_end = 1;
- }
-
- for ( ; ; )
- {
- sp = pending_lapic_eoi_sp(cpu);
- if ( (sp == 0) || !pending_lapic_eoi[cpu][sp-1].ready_to_end )
- {
- spin_unlock_irqrestore(&desc->lock, flags);
- return;
- }
- if ( pending_lapic_eoi[cpu][sp-1].vector != vector )
- {
- spin_unlock(&desc->lock);
- vector = pending_lapic_eoi[cpu][sp-1].vector;
- desc = &irq_desc[vector];
- spin_lock(&desc->lock);
- }
- desc->handler->end(vector);
- pending_lapic_eoi_sp(cpu) = sp-1;
- }
+ if ( !(desc->status & IRQ_GUEST) ||
+ (action->in_flight != 0) ||
+ !test_and_clear_bit(cpu, &action->cpu_eoi_map) )
+ return;
+
+ sp = pending_eoi_sp(cpu);
+ do {
+ ASSERT(sp > 0);
+ } while ( pending_eoi[cpu][--sp].vector != vector );
+ ASSERT(!pending_eoi[cpu][sp].ready);
+ pending_eoi[cpu][sp].ready = 1;
+}
+
+/* Mark specified IRQ as ready-for-EOI (if it really is) and attempt to EOI. */
+static void set_eoi_ready(void *data)
+{
+ irq_desc_t *desc = data;
+
+ ASSERT(!local_irq_is_enabled());
+
+ spin_lock(&desc->lock);
+ __set_eoi_ready(desc);
+ spin_unlock(&desc->lock);
+
+ flush_ready_eoi(NULL);
+}
+
+/*
+ * Forcibly flush all pending EOIs on this CPU by emulating end-of-ISR
+ * notifications from guests. The caller of this function must ensure that
+ * all CPUs execute flush_ready_eoi().
+ */
+static void flush_all_pending_eoi(void *unused)
+{
+ irq_desc_t *desc;
+ irq_guest_action_t *action;
+ int i, vector, sp, cpu = smp_processor_id();
+
+ ASSERT(!local_irq_is_enabled());
+
+ sp = pending_eoi_sp(cpu);
+ while ( --sp >= 0 )
+ {
+ if ( pending_eoi[cpu][sp].ready )
+ continue;
+ vector = pending_eoi[cpu][sp].vector;
+ desc = &irq_desc[vector];
+ spin_lock(&desc->lock);
+ action = (irq_guest_action_t *)desc->action;
+ ASSERT(action->ack_type == ACKTYPE_EOI);
+ ASSERT(desc->status & IRQ_GUEST);
+ for ( i = 0; i < action->nr_guests; i++ )
+ clear_bit(vector_to_irq(vector), &action->guest[i]->pirq_mask);
+ action->in_flight = 0;
+ spin_unlock(&desc->lock);
+ }
+
+ flush_ready_eoi(NULL);
}
int pirq_guest_unmask(struct domain *d)
@@ -262,6 +309,7 @@
action = (irq_guest_action_t *)desc->action;
spin_lock_irq(&desc->lock);
+
if ( !test_bit(d->pirq_to_evtchn[pirq], &s->evtchn_mask[0]) &&
test_and_clear_bit(pirq, &d->pirq_mask) )
{
@@ -273,14 +321,22 @@
cpu_eoi_map = action->cpu_eoi_map;
}
}
- spin_unlock_irq(&desc->lock);
if ( __test_and_clear_bit(cpu, &cpu_eoi_map) )
- end_guest_irq(desc);
+ {
+ __set_eoi_ready(desc);
+ spin_unlock(&desc->lock);
+ flush_ready_eoi(NULL);
+ local_irq_enable();
+ }
+ else
+ {
+ spin_unlock_irq(&desc->lock);
+ }
if ( !cpus_empty(cpu_eoi_map) )
{
- on_selected_cpus(cpu_eoi_map, end_guest_irq, desc, 1, 0);
+ on_selected_cpus(cpu_eoi_map, set_eoi_ready, desc, 1, 0);
cpu_eoi_map = CPU_MASK_NONE;
}
}
@@ -316,7 +372,7 @@
* on which they were received. This is because we tickle the LAPIC to EOI.
*/
if ( !strcmp(desc->handler->typename, "IO-APIC-level") )
- return ioapic_ack_new ? ACKTYPE_LAPIC_EOI : ACKTYPE_UNMASK;
+ return ioapic_ack_new ? ACKTYPE_EOI : ACKTYPE_UNMASK;
BUG();
return 0;
@@ -334,6 +390,7 @@
if ( (irq < 0) || (irq >= NR_IRQS) )
return -EINVAL;
+ retry:
vector = irq_to_vector(irq);
if ( vector == 0 )
return -EINVAL;
@@ -384,6 +441,18 @@
irq);
rc = -EBUSY;
goto out;
+ }
+ else if ( action->nr_guests == 0 )
+ {
+ /*
+ * Indicates that an ACKTYPE_EOI interrupt is being released.
+ * Wait for that to happen before continuing.
+ */
+ ASSERT(action->ack_type == ACKTYPE_EOI);
+ ASSERT(desc->status & IRQ_DISABLED);
+ spin_unlock_irqrestore(&desc->lock, flags);
+ cpu_relax();
+ goto retry;
}
if ( action->nr_guests == IRQ_MAX_GUESTS )
@@ -428,27 +497,16 @@
(--action->in_flight == 0) )
desc->handler->end(vector);
break;
- case ACKTYPE_LAPIC_EOI:
- if ( test_and_clear_bit(irq, &d->pirq_mask) )
- --action->in_flight;
- while ( action->in_flight == 0 )
- {
- /* We cannot release guest info until all pending ACKs are done. */
+ case ACKTYPE_EOI:
+ /* NB. If #guests == 0 then we clear the eoi_map later on. */
+ if ( test_and_clear_bit(irq, &d->pirq_mask) &&
+ (--action->in_flight == 0) &&
+ (action->nr_guests != 0) )
+ {
cpu_eoi_map = action->cpu_eoi_map;
- if ( cpus_empty(cpu_eoi_map) )
- break;
-
- /* We cannot hold the lock while interrupting other CPUs. */
spin_unlock_irqrestore(&desc->lock, flags);
- on_selected_cpus(cpu_eoi_map, end_guest_irq, desc, 1, 1);
+ on_selected_cpus(cpu_eoi_map, set_eoi_ready, desc, 1, 0);
spin_lock_irqsave(&desc->lock, flags);
-
- /* The world can change while we do not hold the lock. */
- if ( !(desc->status & IRQ_GUEST) )
- goto out;
- if ( (action->ack_type != ACKTYPE_LAPIC_EOI) ||
- (action->nr_guests != 0) )
- break;
}
break;
}
@@ -459,12 +517,31 @@
goto out;
BUG_ON(action->in_flight != 0);
+
+ /* Disabling IRQ before releasing the desc_lock avoids an IRQ storm. */
+ desc->depth = 1;
+ desc->status |= IRQ_DISABLED;
+ desc->handler->disable(vector);
+
+ /*
+ * We may have a EOI languishing anywhere in one of the per-CPU
+ * EOI stacks. Forcibly flush the stack on every CPU where this might
+ * be the case.
+ */
+ cpu_eoi_map = action->cpu_eoi_map;
+ if ( !cpus_empty(cpu_eoi_map) )
+ {
+ BUG_ON(action->ack_type != ACKTYPE_EOI);
+ spin_unlock_irqrestore(&desc->lock, flags);
+ on_selected_cpus(cpu_eoi_map, flush_all_pending_eoi, NULL, 1, 1);
+ on_selected_cpus(cpu_online_map, flush_ready_eoi, NULL, 1, 1);
+ spin_lock_irqsave(&desc->lock, flags);
+ }
+
BUG_ON(!cpus_empty(action->cpu_eoi_map));
desc->action = NULL;
xfree(action);
- desc->depth = 1;
- desc->status |= IRQ_DISABLED;
desc->status &= ~IRQ_GUEST;
desc->handler->shutdown(vector);
@@ -543,40 +620,20 @@
static struct timer end_irq_timer[NR_CPUS];
+/*
+ * force_intack: Forcibly emit all pending EOIs on each CPU every second.
+ * Mainly useful for debugging or poking lazy guests ISRs.
+ */
+
static void end_irq_timeout(void *unused)
{
- irq_desc_t *desc;
- irq_guest_action_t *action;
- cpumask_t cpu_eoi_map;
- unsigned int cpu = smp_processor_id();
- int sp, vector, i;
+ int cpu = smp_processor_id();
local_irq_disable();
-
- if ( (sp = pending_lapic_eoi_sp(cpu)) == 0 )
- {
- local_irq_enable();
- return;
- }
-
- vector = pending_lapic_eoi[cpu][sp-1].vector;
- ASSERT(!pending_lapic_eoi[cpu][sp-1].ready_to_end);
-
- desc = &irq_desc[vector];
- spin_lock(&desc->lock);
- action = (irq_guest_action_t *)desc->action;
- ASSERT(action->ack_type == ACKTYPE_LAPIC_EOI);
- ASSERT(desc->status & IRQ_GUEST);
- for ( i = 0; i < action->nr_guests; i++ )
- clear_bit(vector_to_irq(vector), &action->guest[i]->pirq_mask);
- action->in_flight = 0;
- cpu_eoi_map = action->cpu_eoi_map;
- spin_unlock(&desc->lock);
-
+ flush_all_pending_eoi(NULL);
local_irq_enable();
- if ( !cpus_empty(cpu_eoi_map) )
- on_selected_cpus(cpu_eoi_map, end_guest_irq, desc, 1, 0);
+ on_selected_cpus(cpu_online_map, flush_ready_eoi, NULL, 1, 0);
set_timer(&end_irq_timer[cpu], NOW() + MILLISECS(1000));
}
_______________________________________________
Xen-changelog mailing list
Xen-changelog@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-changelog
|
 |
Lists.xenproject.org is hosted with RackSpace, monitoring our |