From 8db6d888e3eb131900111506b93f6101413df5b4 Mon Sep 17 00:00:00 2001
From: zhanghailiang <zhang.zhanghailiang@huawei.com>
Date: Mon, 5 Aug 2019 15:30:05 +0800
Subject: [PATCH] hw/arm64: add vcpu cache info support

Support VCPU Cache info by dtb and PPTT table, including L1, L2 and L3 Cache.

Signed-off-by: zhanghailiang <zhang.zhanghailiang@huawei.com>
Signed-off-by: Honghao <honghao5@huawei.com>
---
 hw/acpi/aml-build.c         | 124 ++++++++++++++++++++++++++++++++++++
 hw/arm/virt.c               |  76 +++++++++++++++++++++-
 include/hw/acpi/aml-build.h |  46 +++++++++++++
 3 files changed, 245 insertions(+), 1 deletion(-)

diff --git a/hw/acpi/aml-build.c b/hw/acpi/aml-build.c
index 9d39ad10..99209c0a 100644
--- a/hw/acpi/aml-build.c
+++ b/hw/acpi/aml-build.c
@@ -55,6 +55,129 @@ static void build_append_array(GArray *array, GArray *val)
 /*
  * ACPI 6.2 Processor Properties Topology Table (PPTT)
  */
+#ifdef __aarch64__
+static void build_cache_head(GArray *tbl, uint32_t next_level)
+{
+    build_append_byte(tbl, 1);
+    build_append_byte(tbl, 24);
+    build_append_int_noprefix(tbl, 0, 2);
+    build_append_int_noprefix(tbl, 127, 4);
+    build_append_int_noprefix(tbl, next_level, 4);
+}
+
+static void build_cache_tail(GArray *tbl, uint32_t cache_type)
+{
+    switch (cache_type) {
+    case ARM_L1D_CACHE: /* L1 dcache info*/
+        build_append_int_noprefix(tbl, ARM_L1DCACHE_SIZE, 4);
+        build_append_int_noprefix(tbl, ARM_L1DCACHE_SET, 4);
+        build_append_byte(tbl, ARM_L1DCACHE_ASSOCIATIVITY);
+        build_append_byte(tbl, ARM_L1DCACHE_ATTRIBUTES);
+        build_append_int_noprefix(tbl, ARM_L1DCACHE_LINE_SIZE, 2);
+        break;
+    case ARM_L1I_CACHE: /* L1 icache info*/
+        build_append_int_noprefix(tbl, ARM_L1ICACHE_SIZE, 4);
+        build_append_int_noprefix(tbl, ARM_L1ICACHE_SET, 4);
+        build_append_byte(tbl, ARM_L1ICACHE_ASSOCIATIVITY);
+        build_append_byte(tbl, ARM_L1ICACHE_ATTRIBUTES);
+        build_append_int_noprefix(tbl, ARM_L1ICACHE_LINE_SIZE, 2);
+        break;
+    case ARM_L2_CACHE: /* L2 cache info*/
+        build_append_int_noprefix(tbl, ARM_L2CACHE_SIZE, 4);
+        build_append_int_noprefix(tbl, ARM_L2CACHE_SET, 4);
+        build_append_byte(tbl, ARM_L2CACHE_ASSOCIATIVITY);
+        build_append_byte(tbl, ARM_L2CACHE_ATTRIBUTES);
+        build_append_int_noprefix(tbl, ARM_L2CACHE_LINE_SIZE, 2);
+        break;
+    case ARM_L3_CACHE: /* L3 cache info*/
+        build_append_int_noprefix(tbl, ARM_L3CACHE_SIZE, 4);
+        build_append_int_noprefix(tbl, ARM_L3CACHE_SET, 4);
+        build_append_byte(tbl, ARM_L3CACHE_ASSOCIATIVITY);
+        build_append_byte(tbl, ARM_L3CACHE_ATTRIBUTES);
+        build_append_int_noprefix(tbl, ARM_L3CACHE_LINE_SIZE, 2);
+        break;
+    default:
+        build_append_int_noprefix(tbl, 0, 4);
+        build_append_int_noprefix(tbl, 0, 4);
+        build_append_byte(tbl, 0);
+        build_append_byte(tbl, 0);
+        build_append_int_noprefix(tbl, 0, 2);
+        break;
+    }
+}
+
+static void build_cache_hierarchy(GArray *tbl,
+              uint32_t next_level, uint32_t cache_type)
+{
+    build_cache_head(tbl, next_level);
+    build_cache_tail(tbl, cache_type);
+}
+
+static void build_arm_socket_hierarchy(GArray *tbl,
+                 uint32_t offset, uint32_t id)
+{
+    build_append_byte(tbl, 0);            /* Type 0 - processor */
+    build_append_byte(tbl, 24);           /* Length, add private resources */
+    build_append_int_noprefix(tbl, 0, 2); /* Reserved */
+    build_append_int_noprefix(tbl, 1, 4); /* Processor boundary and id invalid*/
+    build_append_int_noprefix(tbl, 0, 4);
+    build_append_int_noprefix(tbl, id, 4);
+    build_append_int_noprefix(tbl, 1, 4); /* Num private resources */
+    build_append_int_noprefix(tbl, offset, 4);
+}
+
+static void build_arm_cpu_hierarchy(GArray *tbl,
+                 struct offset_status *offset, uint32_t id)
+{
+    if (!offset) {
+        return;
+    }
+    build_append_byte(tbl, 0);            /* Type 0 - processor */
+    build_append_byte(tbl, 32);           /* Length, add private resources */
+    build_append_int_noprefix(tbl, 0, 2); /* Reserved */
+    build_append_int_noprefix(tbl, 2, 4); /* Valid id*/
+    build_append_int_noprefix(tbl, offset->parent, 4);
+    build_append_int_noprefix(tbl, id, 4);
+    build_append_int_noprefix(tbl, 3, 4); /* Num private resources */
+    build_append_int_noprefix(tbl, offset->l1d_offset, 4);
+    build_append_int_noprefix(tbl, offset->l1i_offset, 4);
+    build_append_int_noprefix(tbl, offset->l2_offset, 4);
+}
+
+void build_pptt(GArray *table_data, BIOSLinker *linker, int possible_cpus)
+{
+    int pptt_start = table_data->len;
+    int uid = 0, cpus = 0, socket;
+    struct offset_status offset;
+
+    acpi_data_push(table_data, sizeof(AcpiTableHeader));
+
+    for (socket = 0; cpus < possible_cpus; socket++) {
+        int core;
+        uint32_t l3_offset = table_data->len - pptt_start;
+        build_cache_hierarchy(table_data, 0, ARM_L3_CACHE);
+
+        offset.parent = table_data->len - pptt_start;
+        build_arm_socket_hierarchy(table_data, l3_offset, socket);
+
+        for (core = 0; core < smp_cores; core++) {
+            offset.l2_offset = table_data->len - pptt_start;
+            build_cache_hierarchy(table_data, 0, ARM_L2_CACHE);
+            offset.l1d_offset = table_data->len - pptt_start;
+            build_cache_hierarchy(table_data, offset.l2_offset, ARM_L1D_CACHE);
+            offset.l1i_offset = table_data->len - pptt_start;
+            build_cache_hierarchy(table_data, offset.l2_offset, ARM_L1I_CACHE);
+            build_arm_cpu_hierarchy(table_data, &offset, uid++);
+            cpus++;
+        }
+    }
+
+    build_header(linker, table_data,
+                 (void *)(table_data->data + pptt_start), "PPTT",
+                 table_data->len - pptt_start, 1, NULL, NULL);
+}
+
+#else
 static void build_cpu_hierarchy(GArray *tbl, uint32_t flags,
                                 uint32_t parent, uint32_t id)
 {
@@ -100,6 +223,7 @@ void build_pptt(GArray *table_data, BIOSLinker *linker, int possible_cpus)
                  (void *)(table_data->data + pptt_start), "PPTT",
                  table_data->len - pptt_start, 1, NULL, NULL);
 }
+#endif
 
 #define ACPI_NAMESEG_LEN 4
 
diff --git a/hw/arm/virt.c b/hw/arm/virt.c
index 16700a2e..96f56e2e 100644
--- a/hw/arm/virt.c
+++ b/hw/arm/virt.c
@@ -304,6 +304,77 @@ static void fdt_add_timer_nodes(const VirtMachineState *vms)
                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
 }
 
+static void fdt_add_l3cache_nodes(const VirtMachineState *vms)
+{
+    int i;
+    unsigned int sockets = vms->smp_cpus / smp_cores;
+
+    /* If current is not equal to max */
+    if (vms->smp_cpus % smp_cores)
+	    sockets++;
+
+    for (i = 0; i < sockets; i++) {
+        char *nodename = g_strdup_printf("/cpus/l3-cache%d", i);
+        qemu_fdt_add_subnode(vms->fdt, nodename);
+        qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cache");
+        qemu_fdt_setprop_string(vms->fdt, nodename, "cache-unified", "true");
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "cache-level", 3);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "cache-size", 0x2000000);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "cache-line-size", 128);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "cache-sets", 2048);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle",
+                              qemu_fdt_alloc_phandle(vms->fdt));
+        g_free(nodename);
+    }
+}
+
+
+static void fdt_add_l2cache_nodes(const VirtMachineState *vms)
+{
+    int i, j;
+    signed int sockets = vms->smp_cpus / smp_cores;
+
+    /* If current is not equal to max */
+    if (vms->smp_cpus % smp_cores)
+	    sockets++;
+
+    for (i = 0; i < sockets; i++) {
+        char *next_path = g_strdup_printf("/cpus/l3-cache%d", i);
+        for (j = 0; j < smp_cores; j++) {
+            char *nodename = g_strdup_printf("/cpus/l2-cache%d",
+                                  i * smp_cores + j);
+            qemu_fdt_add_subnode(vms->fdt, nodename);
+            qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cache");
+            qemu_fdt_setprop_cell(vms->fdt, nodename, "cache-size", 0x80000);
+            qemu_fdt_setprop_cell(vms->fdt, nodename, "cache-line-size", 64);
+            qemu_fdt_setprop_cell(vms->fdt, nodename, "cache-sets", 1024);
+            qemu_fdt_setprop_phandle(vms->fdt, nodename,
+                                  "next-level-cache", next_path);
+            qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle",
+                                  qemu_fdt_alloc_phandle(vms->fdt));
+            g_free(nodename);
+        }
+        g_free(next_path);
+    }
+}
+
+static void fdt_add_l1cache_prop(const VirtMachineState *vms,
+                            char *nodename, int cpu)
+{
+        char *cachename = g_strdup_printf("/cpus/l2-cache%d", cpu);
+
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "d-cache-size", 0x10000);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "d-cache-line-size", 64);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "d-cache-sets", 256);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "i-cache-size", 0x10000);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "i-cache-line-size", 64);
+        qemu_fdt_setprop_cell(vms->fdt, nodename, "i-cache-sets", 256);
+        qemu_fdt_setprop_phandle(vms->fdt, nodename,
+                     "next-level-cache", cachename);
+        g_free(cachename);
+}
+
+
 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
 {
     int cpu;
@@ -336,6 +407,9 @@ static void fdt_add_cpu_nodes(const VirtMachineState *vms)
     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
 
+    fdt_add_l3cache_nodes(vms);
+    fdt_add_l2cache_nodes(vms);
+
     for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
@@ -364,7 +438,7 @@ static void fdt_add_cpu_nodes(const VirtMachineState *vms)
             qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
         }
-
+        fdt_add_l1cache_prop(vms, nodename, cpu);
         qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle",
                               qemu_fdt_alloc_phandle(vms->fdt));
 
diff --git a/include/hw/acpi/aml-build.h b/include/hw/acpi/aml-build.h
index bfb0b100..0be3453a 100644
--- a/include/hw/acpi/aml-build.h
+++ b/include/hw/acpi/aml-build.h
@@ -223,6 +223,52 @@ struct AcpiBuildTables {
     BIOSLinker *linker;
 } AcpiBuildTables;
 
+#ifdef __aarch64__
+/* Definitions of the hardcoded cache info*/
+
+typedef enum {
+    ARM_L1D_CACHE,
+    ARM_L1I_CACHE,
+    ARM_L2_CACHE,
+    ARM_L3_CACHE
+} ArmCacheType;
+
+/* L1 data cache: */
+#define ARM_L1DCACHE_SIZE 65536
+#define ARM_L1DCACHE_SET 256
+#define ARM_L1DCACHE_ASSOCIATIVITY 4
+#define ARM_L1DCACHE_ATTRIBUTES 2
+#define ARM_L1DCACHE_LINE_SIZE 64
+
+/* L1 instruction cache: */
+#define ARM_L1ICACHE_SIZE 65536
+#define ARM_L1ICACHE_SET 256
+#define ARM_L1ICACHE_ASSOCIATIVITY 4
+#define ARM_L1ICACHE_ATTRIBUTES 4
+#define ARM_L1ICACHE_LINE_SIZE 64
+
+/* Level 2 unified cache: */
+#define ARM_L2CACHE_SIZE 524288
+#define ARM_L2CACHE_SET 1024
+#define ARM_L2CACHE_ASSOCIATIVITY 8
+#define ARM_L2CACHE_ATTRIBUTES 10
+#define ARM_L2CACHE_LINE_SIZE 64
+
+/* Level 3 unified cache: */
+#define ARM_L3CACHE_SIZE 33554432
+#define ARM_L3CACHE_SET 2048
+#define ARM_L3CACHE_ASSOCIATIVITY 15
+#define ARM_L3CACHE_ATTRIBUTES 10
+#define ARM_L3CACHE_LINE_SIZE 128
+
+struct offset_status {
+    uint32_t parent;
+    uint32_t l2_offset;
+    uint32_t l1d_offset;
+    uint32_t l1i_offset;
+};
+
+#endif
 /**
  * init_aml_allocator:
  *
-- 
2.23.0