irqchip/gic-v4.1: Properly lock VPEs when doing a directLPI invalidation

We normally rely on the irq_to_cpuid_[un]lock() primitives to make
sure nothing will change col->idx while performing a LPI invalidation.

However, these primitives do not cover VPE doorbells, and we have
some open-coded locking for that. Unfortunately, this locking is
pretty bogus.

Instead, extend the above primitives to cover VPE doorbells and
convert the whole thing to it.

Fixes: f3a05921 ("irqchip/gic-v4.1: Ensure mutual exclusion between vPE affinity change and RD access")
Reported-by: Kunkun Jiang <jiangkunkun@huawei.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: Zenghui Yu <yuzenghui@huawei.com>
Cc: wanghaibin.wang@huawei.com
Tested-by: Kunkun Jiang <jiangkunkun@huawei.com>
Reviewed-by: Zenghui Yu <yuzenghui@huawei.com>
Link: https://lore.kernel.org/r/20230617073242.3199746-1-maz@kernel.org

drivers/irqchip/irq-gic-v3-its.c

@@ -273,13 +273,23 @@ static void vpe_to_cpuid_unlock(struct its_vpe *vpe, unsigned long flags)
 	raw_spin_unlock_irqrestore(&vpe->vpe_lock, flags);
 }
 
+static struct irq_chip its_vpe_irq_chip;
+
 static int irq_to_cpuid_lock(struct irq_data *d, unsigned long *flags)
 {
-	struct its_vlpi_map *map = get_vlpi_map(d);
+	struct its_vpe *vpe = NULL;
 	int cpu;
 
-	if (map) {
-		cpu = vpe_to_cpuid_lock(map->vpe, flags);
+	if (d->chip == &its_vpe_irq_chip) {
+		vpe = irq_data_get_irq_chip_data(d);
+	} else {
+		struct its_vlpi_map *map = get_vlpi_map(d);
+		if (map)
+			vpe = map->vpe;
+	}
+
+	if (vpe) {
+		cpu = vpe_to_cpuid_lock(vpe, flags);
 	} else {
 		/* Physical LPIs are already locked via the irq_desc lock */
 		struct its_device *its_dev = irq_data_get_irq_chip_data(d);
@@ -293,10 +303,18 @@ static int irq_to_cpuid_lock(struct irq_data *d, unsigned long *flags)
 
 static void irq_to_cpuid_unlock(struct irq_data *d, unsigned long flags)
 {
-	struct its_vlpi_map *map = get_vlpi_map(d);
+	struct its_vpe *vpe = NULL;
+
+	if (d->chip == &its_vpe_irq_chip) {
+		vpe = irq_data_get_irq_chip_data(d);
+	} else {
+		struct its_vlpi_map *map = get_vlpi_map(d);
+		if (map)
+			vpe = map->vpe;
+	}
 
-	if (map)
-		vpe_to_cpuid_unlock(map->vpe, flags);
+	if (vpe)
+		vpe_to_cpuid_unlock(vpe, flags);
 }
 
 static struct its_collection *valid_col(struct its_collection *col)
@@ -1433,14 +1451,29 @@ static void wait_for_syncr(void __iomem *rdbase)
 		cpu_relax();
 }
 
-static void direct_lpi_inv(struct irq_data *d)
+static void __direct_lpi_inv(struct irq_data *d, u64 val)
 {
-	struct its_vlpi_map *map = get_vlpi_map(d);
 	void __iomem *rdbase;
 	unsigned long flags;
-	u64 val;
 	int cpu;
 
+	/* Target the redistributor this LPI is currently routed to */
+	cpu = irq_to_cpuid_lock(d, &flags);
+	raw_spin_lock(&gic_data_rdist_cpu(cpu)->rd_lock);
+
+	rdbase = per_cpu_ptr(gic_rdists->rdist, cpu)->rd_base;
+	gic_write_lpir(val, rdbase + GICR_INVLPIR);
+	wait_for_syncr(rdbase);
+
+	raw_spin_unlock(&gic_data_rdist_cpu(cpu)->rd_lock);
+	irq_to_cpuid_unlock(d, flags);
+}
+
+static void direct_lpi_inv(struct irq_data *d)
+{
+	struct its_vlpi_map *map = get_vlpi_map(d);
+	u64 val;
+
 	if (map) {
 		struct its_device *its_dev = irq_data_get_irq_chip_data(d);
 
@@ -1453,15 +1486,7 @@ static void direct_lpi_inv(struct irq_data *d)
 		val = d->hwirq;
 	}
 
-	/* Target the redistributor this LPI is currently routed to */
-	cpu = irq_to_cpuid_lock(d, &flags);
-	raw_spin_lock(&gic_data_rdist_cpu(cpu)->rd_lock);
-	rdbase = per_cpu_ptr(gic_rdists->rdist, cpu)->rd_base;
-	gic_write_lpir(val, rdbase + GICR_INVLPIR);
-
-	wait_for_syncr(rdbase);
-	raw_spin_unlock(&gic_data_rdist_cpu(cpu)->rd_lock);
-	irq_to_cpuid_unlock(d, flags);
+	__direct_lpi_inv(d, val);
 }
 
 static void lpi_update_config(struct irq_data *d, u8 clr, u8 set)
@@ -3953,18 +3978,10 @@ static void its_vpe_send_inv(struct irq_data *d)
 {
 	struct its_vpe *vpe = irq_data_get_irq_chip_data(d);
 
-	if (gic_rdists->has_direct_lpi) {
-		void __iomem *rdbase;
-
-		/* Target the redistributor this VPE is currently known on */
-		raw_spin_lock(&gic_data_rdist_cpu(vpe->col_idx)->rd_lock);
-		rdbase = per_cpu_ptr(gic_rdists->rdist, vpe->col_idx)->rd_base;
-		gic_write_lpir(d->parent_data->hwirq, rdbase + GICR_INVLPIR);
-		wait_for_syncr(rdbase);
-		raw_spin_unlock(&gic_data_rdist_cpu(vpe->col_idx)->rd_lock);
-	} else {
+	if (gic_rdists->has_direct_lpi)
+		__direct_lpi_inv(d, d->parent_data->hwirq);
+	else
 		its_vpe_send_cmd(vpe, its_send_inv);
-	}
 }
 
 static void its_vpe_mask_irq(struct irq_data *d)