[PULL 14/32] hw/arm: add static NVDIMMs in device tree

[Peter Maydell <peter.maydell@linaro.org>]

From: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>

NVDIMM is used for fast rootfs with EROFS, for example by kata
containers. To allow booting with static NVDIMM memory, add them to the
device tree in arm virt machine.

This allows users to boot directly with nvdimm memory devices without
having to rely on ACPI and hotplug.

Verified to work with command invocation:

./qemu-system-aarch64 \
  -M virt,nvdimm=on \
  -cpu cortex-a57 \
  -m 4G,slots=2,maxmem=8G \
  -object memory-backend-file,id=mem1,share=on,mem-path=/tmp/nvdimm,size=4G,readonly=off \
  -device nvdimm,id=nvdimm1,memdev=mem1,unarmed=off \
  -drive file=./debian-12-nocloud-arm64-commited.qcow2,format=qcow2 \
  -kernel ./vmlinuz-6.1.0-13-arm64 \
  -append "root=/dev/vda1 console=ttyAMA0,115200 acpi=off"
  -initrd ./initrd.img-6.1.0-13-arm64 \
  -nographic \
  -serial mon:stdio

Signed-off-by: Manos Pitsidianakis <manos.pitsidianakis@linaro.org>
Message-id: 20250807-nvdimm_arm64_virt-v2-1-b8054578bea8@linaro.org
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
Signed-off-by: Peter Maydell <peter.maydell@linaro.org>
---
 hw/arm/boot.c | 42 ++++++++++++++++++++++++++++++++++++++++++
 hw/arm/virt.c |  8 +++++---
 2 files changed, 47 insertions(+), 3 deletions(-)

diff --git a/hw/arm/boot.c b/hw/arm/boot.c
index d391cd01bb1..1e57c4ab9ee 100644
--- a/hw/arm/boot.c
+++ b/hw/arm/boot.c
@@ -25,6 +25,7 @@
 #include "hw/boards.h"
 #include "system/reset.h"
 #include "hw/loader.h"
+#include "hw/mem/memory-device.h"
 #include "elf.h"
 #include "system/device_tree.h"
 #include "qemu/config-file.h"
@@ -515,6 +516,29 @@ static void fdt_add_psci_node(void *fdt, ARMCPU *armcpu)
     qemu_fdt_setprop_cell(fdt, "/psci", "migrate", migrate_fn);
 }
 
+static int fdt_add_pmem_node(void *fdt, uint32_t acells, uint32_t scells,
+                             int64_t mem_base, int64_t size, int64_t node)
+{
+    int ret;
+
+    g_autofree char *nodename = g_strdup_printf("/pmem@%" PRIx64, mem_base);
+
+    qemu_fdt_add_subnode(fdt, nodename);
+    qemu_fdt_setprop_string(fdt, nodename, "compatible", "pmem-region");
+    ret = qemu_fdt_setprop_sized_cells(fdt, nodename, "reg", acells,
+                                       mem_base, scells, size);
+    if (ret) {
+        return ret;
+    }
+
+    if (node >= 0) {
+        return qemu_fdt_setprop_cell(fdt, nodename, "numa-node-id",
+                                     node);
+    }
+
+    return 0;
+}
+
 int arm_load_dtb(hwaddr addr, const struct arm_boot_info *binfo,
                  hwaddr addr_limit, AddressSpace *as, MachineState *ms,
                  ARMCPU *cpu)
@@ -525,6 +549,7 @@ int arm_load_dtb(hwaddr addr, const struct arm_boot_info *binfo,
     unsigned int i;
     hwaddr mem_base, mem_len;
     char **node_path;
+    g_autofree MemoryDeviceInfoList *md_list = NULL;
     Error *err = NULL;
 
     if (binfo->dtb_filename) {
@@ -628,6 +653,23 @@ int arm_load_dtb(hwaddr addr, const struct arm_boot_info *binfo,
         }
     }
 
+    md_list = qmp_memory_device_list();
+    for (MemoryDeviceInfoList *m = md_list; m != NULL; m = m->next) {
+        MemoryDeviceInfo *mi = m->value;
+
+        if (mi->type == MEMORY_DEVICE_INFO_KIND_NVDIMM) {
+            PCDIMMDeviceInfo *di = mi->u.nvdimm.data;
+
+            rc = fdt_add_pmem_node(fdt, acells, scells,
+                                   di->addr, di->size, di->node);
+            if (rc < 0) {
+                fprintf(stderr, "couldn't add NVDIMM /pmem@%"PRIx64" node\n",
+                        di->addr);
+                goto fail;
+            }
+        }
+    }
+
     rc = fdt_path_offset(fdt, "/chosen");
     if (rc < 0) {
         qemu_fdt_add_subnode(fdt, "/chosen");
diff --git a/hw/arm/virt.c b/hw/arm/virt.c
index 9326cfc895f..1e63f40fbec 100644
--- a/hw/arm/virt.c
+++ b/hw/arm/virt.c
@@ -2917,7 +2917,7 @@ static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
     const MachineState *ms = MACHINE(hotplug_dev);
     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
 
-    if (!vms->acpi_dev) {
+    if (!vms->acpi_dev && !(is_nvdimm && !dev->hotplugged)) {
         error_setg(errp,
                    "memory hotplug is not enabled: missing acpi-ged device");
         return;
@@ -2949,8 +2949,10 @@ static void virt_memory_plug(HotplugHandler *hotplug_dev,
         nvdimm_plug(ms->nvdimms_state);
     }
 
-    hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
-                         dev, &error_abort);
+    if (vms->acpi_dev) {
+        hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
+                             dev, &error_abort);
+    }
 }
 
 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
-- 
2.43.0