diff --git a/arch/x86/include/asm/hw_irq.h b/arch/x86/include/asm/hw_irq.h
index d465ece58151..a09c20c606cb 100644
--- a/arch/x86/include/asm/hw_irq.h
+++ b/arch/x86/include/asm/hw_irq.h
@@ -85,6 +85,7 @@ struct irq_alloc_info {
 		struct ioapic_alloc_info	ioapic;
 		struct uv_alloc_info		uv;
 	};
+	unsigned int vector_align;
 };
 
 struct irq_cfg {
diff --git a/arch/x86/kernel/apic/msi.c b/arch/x86/kernel/apic/msi.c
index 7517eb05bdc1..d7b6e94b1ceb 100644
--- a/arch/x86/kernel/apic/msi.c
+++ b/arch/x86/kernel/apic/msi.c
@@ -168,6 +168,8 @@ int pci_msi_prepare(struct irq_domain *domain, struct device *dev, int nvec,
 		arg->flags |= X86_IRQ_ALLOC_CONTIGUOUS_VECTORS;
 	}
 
+	arg->vector_align = to_pci_dev(dev)->align_msi_vector;
+
 	return 0;
 }
 EXPORT_SYMBOL_GPL(pci_msi_prepare);
diff --git a/arch/x86/kernel/apic/vector.c b/arch/x86/kernel/apic/vector.c
index 3e6f6b448f6a..2e984516df12 100644
--- a/arch/x86/kernel/apic/vector.c
+++ b/arch/x86/kernel/apic/vector.c
@@ -30,6 +30,7 @@ struct apic_chip_data {
 	unsigned int		cpu;
 	unsigned int		prev_cpu;
 	unsigned int		irq;
+	unsigned int		vector_align;
 	struct hlist_node	clist;
 	unsigned int		move_in_progress	: 1,
 				is_managed		: 1,
@@ -190,7 +191,8 @@ static int reserve_managed_vector(struct irq_data *irqd)
 
 	raw_spin_lock_irqsave(&vector_lock, flags);
 	apicd->is_managed = true;
-	ret = irq_matrix_reserve_managed(vector_matrix, affmsk);
+	ret = irq_matrix_reserve_managed(vector_matrix, affmsk,
+					 apicd->vector_align);
 	raw_spin_unlock_irqrestore(&vector_lock, flags);
 	trace_vector_reserve_managed(irqd->irq, ret);
 	return ret;
@@ -245,7 +247,8 @@ assign_vector_locked(struct irq_data *irqd, const struct cpumask *dest)
 	if (apicd->move_in_progress || !hlist_unhashed(&apicd->clist))
 		return -EBUSY;
 
-	vector = irq_matrix_alloc(vector_matrix, dest, resvd, &cpu);
+	vector = irq_matrix_alloc(vector_matrix, dest, resvd, &cpu,
+				  apicd->vector_align);
 	trace_vector_alloc(irqd->irq, vector, resvd, vector);
 	if (vector < 0)
 		return vector;
@@ -322,7 +325,7 @@ assign_managed_vector(struct irq_data *irqd, const struct cpumask *dest)
 	if (apicd->vector && cpumask_test_cpu(apicd->cpu, vector_searchmask))
 		return 0;
 	vector = irq_matrix_alloc_managed(vector_matrix, vector_searchmask,
-					  &cpu);
+					  &cpu, apicd->vector_align);
 	trace_vector_alloc_managed(irqd->irq, vector, vector);
 	if (vector < 0)
 		return vector;
@@ -562,6 +565,10 @@ static int x86_vector_alloc_irqs(struct irq_domain *domain, unsigned int virq,
 			goto error;
 		}
 
+		if (info->type == X86_IRQ_ALLOC_TYPE_PCI_MSI
+				|| info->type == X86_IRQ_ALLOC_TYPE_PCI_MSIX)
+			apicd->vector_align = info->vector_align;
+
 		apicd->irq = virq + i;
 		irqd->chip = &lapic_controller;
 		irqd->chip_data = apicd;
diff --git a/include/linux/irq.h b/include/linux/irq.h
index c3eb89606c2b..19cb159292d6 100644
--- a/include/linux/irq.h
+++ b/include/linux/irq.h
@@ -1242,14 +1242,14 @@ struct irq_matrix *irq_alloc_matrix(unsigned int matrix_bits,
 void irq_matrix_online(struct irq_matrix *m);
 void irq_matrix_offline(struct irq_matrix *m);
 void irq_matrix_assign_system(struct irq_matrix *m, unsigned int bit, bool replace);
-int irq_matrix_reserve_managed(struct irq_matrix *m, const struct cpumask *msk);
+int irq_matrix_reserve_managed(struct irq_matrix *m, const struct cpumask *msk, unsigned int align);
 void irq_matrix_remove_managed(struct irq_matrix *m, const struct cpumask *msk);
 int irq_matrix_alloc_managed(struct irq_matrix *m, const struct cpumask *msk,
-				unsigned int *mapped_cpu);
+				unsigned int *mapped_cpu, unsigned int align);
 void irq_matrix_reserve(struct irq_matrix *m);
 void irq_matrix_remove_reserved(struct irq_matrix *m);
 int irq_matrix_alloc(struct irq_matrix *m, const struct cpumask *msk,
-		     bool reserved, unsigned int *mapped_cpu);
+		     bool reserved, unsigned int *mapped_cpu, unsigned int align);
 void irq_matrix_free(struct irq_matrix *m, unsigned int cpu,
 		     unsigned int bit, bool managed);
 void irq_matrix_assign(struct irq_matrix *m, unsigned int bit);
diff --git a/include/linux/pci.h b/include/linux/pci.h
index b8c354e62937..213103612691 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -485,6 +485,7 @@ struct pci_dev {
 #ifdef CONFIG_PCI_MSI
 	void __iomem	*msix_base;
 	raw_spinlock_t	msi_lock;
+	unsigned int	align_msi_vector;
 #endif
 	struct pci_vpd	vpd;
 #ifdef CONFIG_PCIE_DPC
diff --git a/kernel/irq/matrix.c b/kernel/irq/matrix.c
index 1698e77645ac..4db1b022e904 100644
--- a/kernel/irq/matrix.c
+++ b/kernel/irq/matrix.c
@@ -108,14 +108,17 @@ void irq_matrix_offline(struct irq_matrix *m)
 }
 
 static unsigned int matrix_alloc_area(struct irq_matrix *m, struct cpumap *cm,
-				      unsigned int num, bool managed)
+				      unsigned int num, bool managed, unsigned int align)
 {
 	unsigned int area, start = m->alloc_start;
 	unsigned int end = m->alloc_end;
 
+	if (align > 0)
+		align--;
+
 	bitmap_or(m->scratch_map, cm->managed_map, m->system_map, end);
 	bitmap_or(m->scratch_map, m->scratch_map, cm->alloc_map, end);
-	area = bitmap_find_next_zero_area(m->scratch_map, end, start, num, 0);
+	area = bitmap_find_next_zero_area(m->scratch_map, end, start, num, align);
 	if (area >= end)
 		return area;
 	if (managed)
@@ -207,7 +210,7 @@ void irq_matrix_assign_system(struct irq_matrix *m, unsigned int bit,
  * on all CPUs in @msk, but it's not guaranteed that the bits are at the
  * same offset on all CPUs
  */
-int irq_matrix_reserve_managed(struct irq_matrix *m, const struct cpumask *msk)
+int irq_matrix_reserve_managed(struct irq_matrix *m, const struct cpumask *msk, unsigned int align)
 {
 	unsigned int cpu, failed_cpu;
 
@@ -215,7 +218,7 @@ int irq_matrix_reserve_managed(struct irq_matrix *m, const struct cpumask *msk)
 		struct cpumap *cm = per_cpu_ptr(m->maps, cpu);
 		unsigned int bit;
 
-		bit = matrix_alloc_area(m, cm, 1, true);
+		bit = matrix_alloc_area(m, cm, 1, true, align);
 		if (bit >= m->alloc_end)
 			goto cleanup;
 		cm->managed++;
@@ -284,7 +287,7 @@ void irq_matrix_remove_managed(struct irq_matrix *m, const struct cpumask *msk)
  * @mapped_cpu:	Pointer to store the CPU for which the irq was allocated
  */
 int irq_matrix_alloc_managed(struct irq_matrix *m, const struct cpumask *msk,
-			     unsigned int *mapped_cpu)
+			     unsigned int *mapped_cpu, unsigned int align)
 {
 	unsigned int bit, cpu, end;
 	struct cpumap *cm;
@@ -298,9 +301,14 @@ int irq_matrix_alloc_managed(struct irq_matrix *m, const struct cpumask *msk,
 
 	cm = per_cpu_ptr(m->maps, cpu);
 	end = m->alloc_end;
+
+	if (align > 0)
+		align--;
+
 	/* Get managed bit which are not allocated */
 	bitmap_andnot(m->scratch_map, cm->managed_map, cm->alloc_map, end);
-	bit = find_first_bit(m->scratch_map, end);
+	bitmap_complement(m->scratch_map, m->scratch_map, end);
+	bit = bitmap_find_next_zero_area(m->scratch_map, end, 0, 1, align);
 	if (bit >= end)
 		return -ENOSPC;
 	set_bit(bit, cm->alloc_map);
@@ -375,7 +383,7 @@ void irq_matrix_remove_reserved(struct irq_matrix *m)
  * @mapped_cpu: Pointer to store the CPU for which the irq was allocated
  */
 int irq_matrix_alloc(struct irq_matrix *m, const struct cpumask *msk,
-		     bool reserved, unsigned int *mapped_cpu)
+		     bool reserved, unsigned int *mapped_cpu, unsigned int align)
 {
 	unsigned int cpu, bit;
 	struct cpumap *cm;
@@ -392,7 +400,7 @@ int irq_matrix_alloc(struct irq_matrix *m, const struct cpumask *msk,
 		return -ENOSPC;
 
 	cm = per_cpu_ptr(m->maps, cpu);
-	bit = matrix_alloc_area(m, cm, 1, false);
+	bit = matrix_alloc_area(m, cm, 1, false, align);
 	if (bit >= m->alloc_end)
 		return -ENOSPC;
 	cm->allocated++;
@@ -404,7 +412,6 @@ int irq_matrix_alloc(struct irq_matrix *m, const struct cpumask *msk,
 	*mapped_cpu = cpu;
 	trace_irq_matrix_alloc(bit, cpu, m, cm);
 	return bit;
-
 }
 
 /**