x86/vector/msi: Switch to global reservation mode

Devices with many queues allocate a huge number of interrupts and get
assigned a vector for each of them, even if the queues are not active and
the interrupts never requested. This causes problems with the decision
whether the global vector space is sufficient for CPU hot unplug
operations.

Change it to a reservation scheme, which allows overcommitment.

When the interrupt is allocated and initialized the vector assignment
merily updates the reservation request counter in the matrix
allocator. This counter is used to emit warnings when the reservation
exceeds the available vector space, but does not affect CPU offline
operations. Like the managed interrupts the corresponding MSI/DMAR/IOAPIC
entries are directed to the special shutdown vector.

When the interrupt is requested, then the activation code tries to assign a
real vector. If that succeeds the interrupt is started up and functional.

If that fails, then subsequently request_irq() fails with -ENOSPC.

This allows a clear separation of inactive and active modes and simplifies
the final decisions whether the global vector space is sufficient for CPU
offline operations.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Juergen Gross <jgross@suse.com>
Tested-by: Yu Chen <yu.c.chen@intel.com>
Acked-by: Juergen Gross <jgross@suse.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Alok Kataria <akataria@vmware.com>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Rui Zhang <rui.zhang@intel.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Len Brown <lenb@kernel.org>
Link: https://lkml.kernel.org/r/20170913213156.184211133@linutronix.de

arch/x86/kernel/apic/vector.c

@@ -33,7 +33,9 @@ struct apic_chip_data {
 	unsigned int		irq;
 	struct hlist_node	clist;
 	unsigned int		move_in_progress	: 1,
-				is_managed		: 1;
+				is_managed		: 1,
+				can_reserve		: 1,
+				has_reserved		: 1;
 };
 
 struct irq_domain *x86_vector_domain;
@@ -175,9 +177,31 @@ static int reserve_managed_vector(struct irq_data *irqd)
 	return ret;
 }
 
+static void reserve_irq_vector_locked(struct irq_data *irqd)
+{
+	struct apic_chip_data *apicd = apic_chip_data(irqd);
+
+	irq_matrix_reserve(vector_matrix);
+	apicd->can_reserve = true;
+	apicd->has_reserved = true;
+	trace_vector_reserve(irqd->irq, 0);
+	vector_assign_managed_shutdown(irqd);
+}
+
+static int reserve_irq_vector(struct irq_data *irqd)
+{
+	unsigned long flags;
+
+	raw_spin_lock_irqsave(&vector_lock, flags);
+	reserve_irq_vector_locked(irqd);
+	raw_spin_unlock_irqrestore(&vector_lock, flags);
+	return 0;
+}
+
 static int allocate_vector(struct irq_data *irqd, const struct cpumask *dest)
 {
 	struct apic_chip_data *apicd = apic_chip_data(irqd);
+	bool resvd = apicd->has_reserved;
 	unsigned int cpu = apicd->cpu;
 	int vector = apicd->vector;
 
@@ -191,10 +215,10 @@ static int allocate_vector(struct irq_data *irqd, const struct cpumask *dest)
 	if (vector && cpu_online(cpu) && cpumask_test_cpu(cpu, dest))
 		return 0;
 
-	vector = irq_matrix_alloc(vector_matrix, dest, false, &cpu);
+	vector = irq_matrix_alloc(vector_matrix, dest, resvd, &cpu);
 	if (vector > 0)
 		apic_update_vector(irqd, vector, cpu);
-	trace_vector_alloc(irqd->irq, vector, false, vector);
+	trace_vector_alloc(irqd->irq, vector, resvd, vector);
 	return vector;
 }
 
@@ -252,7 +276,11 @@ assign_irq_vector_policy(struct irq_data *irqd, struct irq_alloc_info *info)
 		return reserve_managed_vector(irqd);
 	if (info->mask)
 		return assign_irq_vector(irqd, info->mask);
+	if (info->type != X86_IRQ_ALLOC_TYPE_MSI &&
+	    info->type != X86_IRQ_ALLOC_TYPE_MSIX)
 		return assign_irq_vector_any(irqd);
+	/* For MSI(X) make only a global reservation with no guarantee */
+	return reserve_irq_vector(irqd);
 }
 
 static int
@@ -314,17 +342,35 @@ static void x86_vector_deactivate(struct irq_domain *dom, struct irq_data *irqd)
 	unsigned long flags;
 
 	trace_vector_deactivate(irqd->irq, apicd->is_managed,
-				false, false);
+				apicd->can_reserve, false);
 
-	if (apicd->is_managed)
+	/* Regular fixed assigned interrupt */
+	if (!apicd->is_managed && !apicd->can_reserve)
+		return;
+	/* If the interrupt has a global reservation, nothing to do */
+	if (apicd->has_reserved)
 		return;
 
 	raw_spin_lock_irqsave(&vector_lock, flags);
 	clear_irq_vector(irqd);
+	if (apicd->can_reserve)
+		reserve_irq_vector_locked(irqd);
+	else
 		vector_assign_managed_shutdown(irqd);
 	raw_spin_unlock_irqrestore(&vector_lock, flags);
 }
 
+static int activate_reserved(struct irq_data *irqd)
+{
+	struct apic_chip_data *apicd = apic_chip_data(irqd);
+	int ret;
+
+	ret = assign_irq_vector_any_locked(irqd);
+	if (!ret)
+		apicd->has_reserved = false;
+	return ret;
+}
+
 static int activate_managed(struct irq_data *irqd)
 {
 	const struct cpumask *dest = irq_data_get_affinity_mask(irqd);
@@ -357,16 +403,19 @@ static int x86_vector_activate(struct irq_domain *dom, struct irq_data *irqd,
 	int ret = 0;
 
 	trace_vector_activate(irqd->irq, apicd->is_managed,
-				false, early);
+			      apicd->can_reserve, early);
 
-	if (!apicd->is_managed)
+	/* Nothing to do for fixed assigned vectors */
+	if (!apicd->can_reserve && !apicd->is_managed)
 		return 0;
 
 	raw_spin_lock_irqsave(&vector_lock, flags);
 	if (early || irqd_is_managed_and_shutdown(irqd))
 		vector_assign_managed_shutdown(irqd);
-	else
+	else if (apicd->is_managed)
 		ret = activate_managed(irqd);
+	else if (apicd->has_reserved)
+		ret = activate_reserved(irqd);
 	raw_spin_unlock_irqrestore(&vector_lock, flags);
 	return ret;
 }
@@ -376,8 +425,11 @@ static void vector_free_reserved_and_managed(struct irq_data *irqd)
 	const struct cpumask *dest = irq_data_get_affinity_mask(irqd);
 	struct apic_chip_data *apicd = apic_chip_data(irqd);
 
-	trace_vector_teardown(irqd->irq, apicd->is_managed, false);
+	trace_vector_teardown(irqd->irq, apicd->is_managed,
+			      apicd->has_reserved);
 
+	if (apicd->has_reserved)
+		irq_matrix_remove_reserved(vector_matrix);
 	if (apicd->is_managed)
 		irq_matrix_remove_managed(vector_matrix, dest);
 }
@@ -604,22 +656,6 @@ int __init arch_early_irq_init(void)
 }
 
 #ifdef CONFIG_SMP
-/* Temporary hack to keep things working */
-static void vector_update_shutdown_irqs(void)
-{
-	struct irq_desc *desc;
-	int irq;
-
-	for_each_irq_desc(irq, desc) {
-		struct irq_data *irqd = irq_desc_get_irq_data(desc);
-		struct apic_chip_data *ad = apic_chip_data(irqd);
-
-		if (!ad || !ad->vector || ad->cpu != smp_processor_id())
-			continue;
-		this_cpu_write(vector_irq[ad->vector], desc);
-		irq_matrix_assign(vector_matrix, ad->vector);
-	}
-}
 
 static struct irq_desc *__setup_vector_irq(int vector)
 {
@@ -655,13 +691,6 @@ void lapic_online(void)
 	 */
 	for (vector = 0; vector < NR_VECTORS; vector++)
 		this_cpu_write(vector_irq[vector], __setup_vector_irq(vector));
-
-	/*
-	 * Until the rewrite of the managed interrupt management is in
-	 * place it's necessary to walk the irq descriptors and check for
-	 * interrupts which are targeted at this CPU.
-	 */
-	vector_update_shutdown_irqs();
 }
 
 void lapic_offline(void)