[RESPIN PATCH] irqchip/gic-v4.1: Use local 4_1 ITS to generate VSGI

[Nianyao Tang]

On multi-node GICv4.1 system, VSGI senders always use one certain 4_1 ITS,
because find_4_1_its return the first its_node in list, regardless of
which node the VSGI sender is on. This brings guest vsgi performance drop
when VM is not deployed on the same node as this returned ITS.

On a 2-socket environment, each with one ITS and 32 cpu, GICv4.1 enabled,
4U8G guest, 4 vcpu is deployed on same socket.
When VM on socket0, kvm-unit-tests ipi_hw result is 850ns.
When VM on socket1, it is 750ns. The reason is VSGI sender always
use lasted reported ITS(that on socket1) to inject VSGI. The access
from cpu to other-socket ITS will cost 100ns more compared to cpu to 
local ITS.

To use local ITS, we can get 12% reduction in IPI latency.

The patch modify find_4_1_its to firstly return per-cpu local_4_1_its,
which is init when inherit the VPE table from the ITS on secondary CPUs.
If fail to find local 4_1 ITS, return any 4_1 ITS like before.

Signed-off-by: Nianyao Tang <tangnianyao at huawei.com>
Reviewed-by: Marc Zyngier <maz at kernel.org>
---
 drivers/irqchip/irq-gic-v3-its.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)

diff --git a/drivers/irqchip/irq-gic-v3-its.c b/drivers/irqchip/irq-gic-v3-its.c
index 3c755d5dad6e..f99c0a86320b 100644
--- a/drivers/irqchip/irq-gic-v3-its.c
+++ b/drivers/irqchip/irq-gic-v3-its.c
@@ -119,6 +119,8 @@ struct its_node {
 	int			vlpi_redist_offset;
 };
 
+static DEFINE_PER_CPU(struct its_node *, local_4_1_its);
+
 #define is_v4(its)		(!!((its)->typer & GITS_TYPER_VLPIS))
 #define is_v4_1(its)		(!!((its)->typer & GITS_TYPER_VMAPP))
 #define device_ids(its)		(FIELD_GET(GITS_TYPER_DEVBITS, (its)->typer) + 1)
@@ -2709,6 +2711,8 @@ static u64 inherit_vpe_l1_table_from_its(void)
 		}
 		val |= FIELD_PREP(GICR_VPROPBASER_4_1_SIZE, GITS_BASER_NR_PAGES(baser) - 1);
 
+		*this_cpu_ptr(&local_4_1_its) = its;
+
 		return val;
 	}
 
@@ -2746,6 +2750,8 @@ static u64 inherit_vpe_l1_table_from_rd(cpumask_t **mask)
 		gic_data_rdist()->vpe_l1_base = gic_data_rdist_cpu(cpu)->vpe_l1_base;
 		*mask = gic_data_rdist_cpu(cpu)->vpe_table_mask;
 
+		*this_cpu_ptr(&local_4_1_its) = *per_cpu_ptr(&local_4_1_its, cpu);
+
 		return val;
 	}
 
@@ -4058,8 +4064,9 @@ static struct irq_chip its_vpe_irq_chip = {
 
 static struct its_node *find_4_1_its(void)
 {
-	static struct its_node *its = NULL;
+	struct its_node *its;
 
+	its = *this_cpu_ptr(&local_4_1_its);
 	if (!its) {
 		list_for_each_entry(its, &its_nodes, entry) {
 			if (is_v4_1(its))
-- 
2.30.0