qemu/hw-arm64-add-vcpu-cache-info-support.patch

From 7d3d37d3af4278aee627952d6a81b63dec6ac62b Mon Sep 17 00:00:00 2001
From: Ying Fang <fangying1@huawei.com>
Date: Sun, 17 Mar 2024 18:56:09 +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>
Signed-off-by: Ying Fang <fangying1@huawei.com>
Signed-off-by: Yanan Wang <wangyanan55@huawei.com>
Signed-off-by: Yuan Zhang <zhangyuan162@huawei.com>
---
 hw/acpi/aml-build.c         | 158 ++++++++++++++++++++++++++++++++++++
 hw/arm/virt.c               |  72 ++++++++++++++++
 include/hw/acpi/aml-build.h |  47 +++++++++++
 3 files changed, 277 insertions(+)

diff --git a/hw/acpi/aml-build.c b/hw/acpi/aml-build.c
index af66bde0f5..2968df5562 100644
--- a/hw/acpi/aml-build.c
+++ b/hw/acpi/aml-build.c
@@ -1994,6 +1994,163 @@ static void build_processor_hierarchy_node(GArray *tbl, uint32_t flags,
     }
 }
 
+#ifdef __aarch64__
+/*
+ * ACPI spec, Revision 6.3
+ * 5.2.29.2 Cache Type Structure (Type 1)
+ */
+static void build_cache_hierarchy_node(GArray *tbl, uint32_t next_level,
+                                       uint32_t cache_type)
+{
+    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);
+
+    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_SETS, 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_SETS, 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_SETS, 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_SETS, 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);
+    }
+}
+
+/*
+ * ACPI spec, Revision 6.3
+ * 5.2.29 Processor Properties Topology Table (PPTT)
+ */
+void build_pptt(GArray *table_data, BIOSLinker *linker, MachineState *ms,
+                const char *oem_id, const char *oem_table_id)
+{
+    MachineClass *mc = MACHINE_GET_CLASS(ms);
+    GQueue *list = g_queue_new();
+    guint pptt_start = table_data->len;
+    guint parent_offset;
+    guint length, i;
+    int uid = 0;
+    int socket;
+    AcpiTable table = { .sig = "PPTT", .rev = 2,
+                        .oem_id = oem_id, .oem_table_id = oem_table_id };
+
+    acpi_table_begin(&table, table_data);
+
+    for (socket = 0; socket < ms->smp.sockets; socket++) {
+        uint32_t l3_cache_offset = table_data->len - pptt_start;
+        build_cache_hierarchy_node(table_data, 0, ARM_L3_CACHE);
+
+        g_queue_push_tail(list,
+            GUINT_TO_POINTER(table_data->len - pptt_start));
+        build_processor_hierarchy_node(
+            table_data,
+            /*
+             * Physical package - represents the boundary
+             * of a physical package
+             */
+            (1 << 0),
+            0, socket, &l3_cache_offset, 1);
+    }
+
+    if (mc->smp_props.clusters_supported) {
+        length = g_queue_get_length(list);
+        for (i = 0; i < length; i++) {
+            int cluster;
+
+            parent_offset = GPOINTER_TO_UINT(g_queue_pop_head(list));
+            for (cluster = 0; cluster < ms->smp.clusters; cluster++) {
+                g_queue_push_tail(list,
+                    GUINT_TO_POINTER(table_data->len - pptt_start));
+                build_processor_hierarchy_node(
+                    table_data,
+                    (0 << 0), /* not a physical package */
+                    parent_offset, cluster, NULL, 0);
+            }
+        }
+    }
+
+    length = g_queue_get_length(list);
+    for (i = 0; i < length; i++) {
+        int core;
+
+        parent_offset = GPOINTER_TO_UINT(g_queue_pop_head(list));
+        for (core = 0; core < ms->smp.cores; core++) {
+            uint32_t priv_rsrc[3] = {};
+            priv_rsrc[0] = table_data->len - pptt_start; /* L2 cache offset */
+            build_cache_hierarchy_node(table_data, 0, ARM_L2_CACHE);
+
+            priv_rsrc[1] = table_data->len - pptt_start; /* L1 dcache offset */
+            build_cache_hierarchy_node(table_data, priv_rsrc[0], ARM_L1D_CACHE);
+
+            priv_rsrc[2] = table_data->len - pptt_start; /* L1 icache offset */
+            build_cache_hierarchy_node(table_data, priv_rsrc[0], ARM_L1I_CACHE);
+
+            if (ms->smp.threads > 1) {
+                g_queue_push_tail(list,
+                    GUINT_TO_POINTER(table_data->len - pptt_start));
+                build_processor_hierarchy_node(
+                    table_data,
+                    (0 << 0), /* not a physical package */
+                    parent_offset, core, priv_rsrc, 3);
+            } else {
+                build_processor_hierarchy_node(
+                    table_data,
+                    (1 << 1) | /* ACPI Processor ID valid */
+                    (1 << 3),  /* Node is a Leaf */
+                    parent_offset, uid++, priv_rsrc, 3);
+            }
+        }
+    }
+
+    length = g_queue_get_length(list);
+    for (i = 0; i < length; i++) {
+        int thread;
+
+        parent_offset = GPOINTER_TO_UINT(g_queue_pop_head(list));
+        for (thread = 0; thread < ms->smp.threads; thread++) {
+            build_processor_hierarchy_node(
+                table_data,
+                (1 << 1) | /* ACPI Processor ID valid */
+                (1 << 2) | /* Processor is a Thread */
+                (1 << 3),  /* Node is a Leaf */
+                parent_offset, uid++, NULL, 0);
+        }
+    }
+
+    g_queue_free(list);
+    acpi_table_end(linker, &table);
+}
+
+#else
 /*
  * ACPI spec, Revision 6.3
  * 5.2.29 Processor Properties Topology Table (PPTT)
@@ -2069,6 +2226,7 @@ void build_pptt(GArray *table_data, BIOSLinker *linker, MachineState *ms,
 
     acpi_table_end(linker, &table);
 }
+#endif
 
 /* build rev1/rev3/rev5.1/rev6.0 FADT */
 void build_fadt(GArray *tbl, BIOSLinker *linker, const AcpiFadtData *f,
diff --git a/hw/arm/virt.c b/hw/arm/virt.c
index 500a15aa5b..b82bd1b8c8 100644
--- a/hw/arm/virt.c
+++ b/hw/arm/virt.c
@@ -379,6 +379,72 @@ static void fdt_add_timer_nodes(const VirtMachineState *vms)
                            INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags);
 }
 
+static void fdt_add_l3cache_nodes(const VirtMachineState *vms)
+{
+    int i;
+    const MachineState *ms = MACHINE(vms);
+    int cpus_per_socket = ms->smp.clusters * ms->smp.cores * ms->smp.threads;
+    int sockets = (ms->smp.cpus + cpus_per_socket - 1) / cpus_per_socket;
+
+    for (i = 0; i < sockets; i++) {
+        char *nodename = g_strdup_printf("/cpus/l3-cache%d", i);
+
+        qemu_fdt_add_subnode(ms->fdt, nodename);
+        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cache");
+        qemu_fdt_setprop_string(ms->fdt, nodename, "cache-unified", "true");
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "cache-level", 3);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "cache-size", 0x2000000);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "cache-line-size", 128);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "cache-sets", 2048);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
+                              qemu_fdt_alloc_phandle(ms->fdt));
+        g_free(nodename);
+    }
+}
+
+static void fdt_add_l2cache_nodes(const VirtMachineState *vms)
+{
+    const MachineState *ms = MACHINE(vms);
+    int cpus_per_socket = ms->smp.clusters * ms->smp.cores * ms->smp.threads;
+    int cpu;
+
+    for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
+        char *next_path = g_strdup_printf("/cpus/l3-cache%d",
+                                          cpu / cpus_per_socket);
+        char *nodename = g_strdup_printf("/cpus/l2-cache%d", cpu);
+
+        qemu_fdt_add_subnode(ms->fdt, nodename);
+        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cache");
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "cache-size", 0x80000);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "cache-line-size", 64);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "cache-sets", 1024);
+        qemu_fdt_setprop_phandle(ms->fdt, nodename, "next-level-cache",
+                                 next_path);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
+                              qemu_fdt_alloc_phandle(ms->fdt));
+
+        g_free(next_path);
+        g_free(nodename);
+    }
+}
+
+static void fdt_add_l1cache_prop(const VirtMachineState *vms,
+                                 char *nodename, int cpu)
+{
+        const MachineState *ms = MACHINE(vms);
+        char *cachename = g_strdup_printf("/cpus/l2-cache%d", cpu);
+
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "d-cache-size", 0x10000);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "d-cache-line-size", 64);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "d-cache-sets", 256);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "i-cache-size", 0x10000);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "i-cache-line-size", 64);
+        qemu_fdt_setprop_cell(ms->fdt, nodename, "i-cache-sets", 256);
+        qemu_fdt_setprop_phandle(ms->fdt, nodename, "next-level-cache",
+                                 cachename);
+        g_free(cachename);
+}
+
 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
 {
     int cpu;
@@ -413,6 +479,11 @@ static void fdt_add_cpu_nodes(const VirtMachineState *vms)
     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
 
+    if (!vmc->no_cpu_topology) {
+        fdt_add_l3cache_nodes(vms);
+        fdt_add_l2cache_nodes(vms);
+    }
+
     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
@@ -442,6 +513,7 @@ static void fdt_add_cpu_nodes(const VirtMachineState *vms)
         }
 
         if (!vmc->no_cpu_topology) {
+            fdt_add_l1cache_prop(vms, nodename, cpu);
             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
                                   qemu_fdt_alloc_phandle(ms->fdt));
         }
diff --git a/include/hw/acpi/aml-build.h b/include/hw/acpi/aml-build.h
index ff2a310270..84ded2ecd3 100644
--- a/include/hw/acpi/aml-build.h
+++ b/include/hw/acpi/aml-build.h
@@ -221,6 +221,53 @@ 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_SETS 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_SETS 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_SETS 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_SETS 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
+
 typedef
 struct CrsRangeEntry {
     uint64_t base;
-- 
2.27.0