xref: /openbmc/qemu/hw/riscv/virt.c (revision efb91426)
1 /*
2  * QEMU RISC-V VirtIO Board
3  *
4  * Copyright (c) 2017 SiFive, Inc.
5  *
6  * RISC-V machine with 16550a UART and VirtIO MMIO
7  *
8  * This program is free software; you can redistribute it and/or modify it
9  * under the terms and conditions of the GNU General Public License,
10  * version 2 or later, as published by the Free Software Foundation.
11  *
12  * This program is distributed in the hope it will be useful, but WITHOUT
13  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
14  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
15  * more details.
16  *
17  * You should have received a copy of the GNU General Public License along with
18  * this program.  If not, see <http://www.gnu.org/licenses/>.
19  */
20 
21 #include "qemu/osdep.h"
22 #include "qemu/units.h"
23 #include "qemu/error-report.h"
24 #include "qemu/guest-random.h"
25 #include "qapi/error.h"
26 #include "hw/boards.h"
27 #include "hw/loader.h"
28 #include "hw/sysbus.h"
29 #include "hw/qdev-properties.h"
30 #include "hw/char/serial.h"
31 #include "target/riscv/cpu.h"
32 #include "hw/core/sysbus-fdt.h"
33 #include "target/riscv/pmu.h"
34 #include "hw/riscv/riscv_hart.h"
35 #include "hw/riscv/virt.h"
36 #include "hw/riscv/boot.h"
37 #include "hw/riscv/numa.h"
38 #include "kvm/kvm_riscv.h"
39 #include "hw/intc/riscv_aclint.h"
40 #include "hw/intc/riscv_aplic.h"
41 #include "hw/intc/sifive_plic.h"
42 #include "hw/misc/sifive_test.h"
43 #include "hw/platform-bus.h"
44 #include "chardev/char.h"
45 #include "sysemu/device_tree.h"
46 #include "sysemu/sysemu.h"
47 #include "sysemu/tcg.h"
48 #include "sysemu/kvm.h"
49 #include "sysemu/tpm.h"
50 #include "hw/pci/pci.h"
51 #include "hw/pci-host/gpex.h"
52 #include "hw/display/ramfb.h"
53 #include "hw/acpi/aml-build.h"
54 #include "qapi/qapi-visit-common.h"
55 
56 /* KVM AIA only supports APLIC MSI. APLIC Wired is always emulated by QEMU. */
57 static bool virt_use_kvm_aia(RISCVVirtState *s)
58 {
59     return kvm_irqchip_in_kernel() && s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC;
60 }
61 
62 static const MemMapEntry virt_memmap[] = {
63     [VIRT_DEBUG] =        {        0x0,         0x100 },
64     [VIRT_MROM] =         {     0x1000,        0xf000 },
65     [VIRT_TEST] =         {   0x100000,        0x1000 },
66     [VIRT_RTC] =          {   0x101000,        0x1000 },
67     [VIRT_CLINT] =        {  0x2000000,       0x10000 },
68     [VIRT_ACLINT_SSWI] =  {  0x2F00000,        0x4000 },
69     [VIRT_PCIE_PIO] =     {  0x3000000,       0x10000 },
70     [VIRT_PLATFORM_BUS] = {  0x4000000,     0x2000000 },
71     [VIRT_PLIC] =         {  0xc000000, VIRT_PLIC_SIZE(VIRT_CPUS_MAX * 2) },
72     [VIRT_APLIC_M] =      {  0xc000000, APLIC_SIZE(VIRT_CPUS_MAX) },
73     [VIRT_APLIC_S] =      {  0xd000000, APLIC_SIZE(VIRT_CPUS_MAX) },
74     [VIRT_UART0] =        { 0x10000000,         0x100 },
75     [VIRT_VIRTIO] =       { 0x10001000,        0x1000 },
76     [VIRT_FW_CFG] =       { 0x10100000,          0x18 },
77     [VIRT_FLASH] =        { 0x20000000,     0x4000000 },
78     [VIRT_IMSIC_M] =      { 0x24000000, VIRT_IMSIC_MAX_SIZE },
79     [VIRT_IMSIC_S] =      { 0x28000000, VIRT_IMSIC_MAX_SIZE },
80     [VIRT_PCIE_ECAM] =    { 0x30000000,    0x10000000 },
81     [VIRT_PCIE_MMIO] =    { 0x40000000,    0x40000000 },
82     [VIRT_DRAM] =         { 0x80000000,           0x0 },
83 };
84 
85 /* PCIe high mmio is fixed for RV32 */
86 #define VIRT32_HIGH_PCIE_MMIO_BASE  0x300000000ULL
87 #define VIRT32_HIGH_PCIE_MMIO_SIZE  (4 * GiB)
88 
89 /* PCIe high mmio for RV64, size is fixed but base depends on top of RAM */
90 #define VIRT64_HIGH_PCIE_MMIO_SIZE  (16 * GiB)
91 
92 static MemMapEntry virt_high_pcie_memmap;
93 
94 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
95 
96 static PFlashCFI01 *virt_flash_create1(RISCVVirtState *s,
97                                        const char *name,
98                                        const char *alias_prop_name)
99 {
100     /*
101      * Create a single flash device.  We use the same parameters as
102      * the flash devices on the ARM virt board.
103      */
104     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
105 
106     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
107     qdev_prop_set_uint8(dev, "width", 4);
108     qdev_prop_set_uint8(dev, "device-width", 2);
109     qdev_prop_set_bit(dev, "big-endian", false);
110     qdev_prop_set_uint16(dev, "id0", 0x89);
111     qdev_prop_set_uint16(dev, "id1", 0x18);
112     qdev_prop_set_uint16(dev, "id2", 0x00);
113     qdev_prop_set_uint16(dev, "id3", 0x00);
114     qdev_prop_set_string(dev, "name", name);
115 
116     object_property_add_child(OBJECT(s), name, OBJECT(dev));
117     object_property_add_alias(OBJECT(s), alias_prop_name,
118                               OBJECT(dev), "drive");
119 
120     return PFLASH_CFI01(dev);
121 }
122 
123 static void virt_flash_create(RISCVVirtState *s)
124 {
125     s->flash[0] = virt_flash_create1(s, "virt.flash0", "pflash0");
126     s->flash[1] = virt_flash_create1(s, "virt.flash1", "pflash1");
127 }
128 
129 static void virt_flash_map1(PFlashCFI01 *flash,
130                             hwaddr base, hwaddr size,
131                             MemoryRegion *sysmem)
132 {
133     DeviceState *dev = DEVICE(flash);
134 
135     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
136     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
137     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
138     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
139 
140     memory_region_add_subregion(sysmem, base,
141                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
142                                                        0));
143 }
144 
145 static void virt_flash_map(RISCVVirtState *s,
146                            MemoryRegion *sysmem)
147 {
148     hwaddr flashsize = virt_memmap[VIRT_FLASH].size / 2;
149     hwaddr flashbase = virt_memmap[VIRT_FLASH].base;
150 
151     virt_flash_map1(s->flash[0], flashbase, flashsize,
152                     sysmem);
153     virt_flash_map1(s->flash[1], flashbase + flashsize, flashsize,
154                     sysmem);
155 }
156 
157 static void create_pcie_irq_map(RISCVVirtState *s, void *fdt, char *nodename,
158                                 uint32_t irqchip_phandle)
159 {
160     int pin, dev;
161     uint32_t irq_map_stride = 0;
162     uint32_t full_irq_map[GPEX_NUM_IRQS * GPEX_NUM_IRQS *
163                           FDT_MAX_INT_MAP_WIDTH] = {};
164     uint32_t *irq_map = full_irq_map;
165 
166     /* This code creates a standard swizzle of interrupts such that
167      * each device's first interrupt is based on it's PCI_SLOT number.
168      * (See pci_swizzle_map_irq_fn())
169      *
170      * We only need one entry per interrupt in the table (not one per
171      * possible slot) seeing the interrupt-map-mask will allow the table
172      * to wrap to any number of devices.
173      */
174     for (dev = 0; dev < GPEX_NUM_IRQS; dev++) {
175         int devfn = dev * 0x8;
176 
177         for (pin = 0; pin < GPEX_NUM_IRQS; pin++) {
178             int irq_nr = PCIE_IRQ + ((pin + PCI_SLOT(devfn)) % GPEX_NUM_IRQS);
179             int i = 0;
180 
181             /* Fill PCI address cells */
182             irq_map[i] = cpu_to_be32(devfn << 8);
183             i += FDT_PCI_ADDR_CELLS;
184 
185             /* Fill PCI Interrupt cells */
186             irq_map[i] = cpu_to_be32(pin + 1);
187             i += FDT_PCI_INT_CELLS;
188 
189             /* Fill interrupt controller phandle and cells */
190             irq_map[i++] = cpu_to_be32(irqchip_phandle);
191             irq_map[i++] = cpu_to_be32(irq_nr);
192             if (s->aia_type != VIRT_AIA_TYPE_NONE) {
193                 irq_map[i++] = cpu_to_be32(0x4);
194             }
195 
196             if (!irq_map_stride) {
197                 irq_map_stride = i;
198             }
199             irq_map += irq_map_stride;
200         }
201     }
202 
203     qemu_fdt_setprop(fdt, nodename, "interrupt-map", full_irq_map,
204                      GPEX_NUM_IRQS * GPEX_NUM_IRQS *
205                      irq_map_stride * sizeof(uint32_t));
206 
207     qemu_fdt_setprop_cells(fdt, nodename, "interrupt-map-mask",
208                            0x1800, 0, 0, 0x7);
209 }
210 
211 static void create_fdt_socket_cpus(RISCVVirtState *s, int socket,
212                                    char *clust_name, uint32_t *phandle,
213                                    uint32_t *intc_phandles)
214 {
215     int cpu;
216     uint32_t cpu_phandle;
217     MachineState *ms = MACHINE(s);
218     char *name, *cpu_name, *core_name, *intc_name, *sv_name;
219     bool is_32_bit = riscv_is_32bit(&s->soc[0]);
220     uint8_t satp_mode_max;
221 
222     for (cpu = s->soc[socket].num_harts - 1; cpu >= 0; cpu--) {
223         RISCVCPU *cpu_ptr = &s->soc[socket].harts[cpu];
224 
225         cpu_phandle = (*phandle)++;
226 
227         cpu_name = g_strdup_printf("/cpus/cpu@%d",
228             s->soc[socket].hartid_base + cpu);
229         qemu_fdt_add_subnode(ms->fdt, cpu_name);
230 
231         if (cpu_ptr->cfg.satp_mode.supported != 0) {
232             satp_mode_max = satp_mode_max_from_map(cpu_ptr->cfg.satp_mode.map);
233             sv_name = g_strdup_printf("riscv,%s",
234                                       satp_mode_str(satp_mode_max, is_32_bit));
235             qemu_fdt_setprop_string(ms->fdt, cpu_name, "mmu-type", sv_name);
236             g_free(sv_name);
237         }
238 
239         name = riscv_isa_string(cpu_ptr);
240         qemu_fdt_setprop_string(ms->fdt, cpu_name, "riscv,isa", name);
241         g_free(name);
242 
243         if (cpu_ptr->cfg.ext_zicbom) {
244             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cbom-block-size",
245                                   cpu_ptr->cfg.cbom_blocksize);
246         }
247 
248         if (cpu_ptr->cfg.ext_zicboz) {
249             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cboz-block-size",
250                                   cpu_ptr->cfg.cboz_blocksize);
251         }
252 
253         if (cpu_ptr->cfg.ext_zicbop) {
254             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cbop-block-size",
255                                   cpu_ptr->cfg.cbop_blocksize);
256         }
257 
258         qemu_fdt_setprop_string(ms->fdt, cpu_name, "compatible", "riscv");
259         qemu_fdt_setprop_string(ms->fdt, cpu_name, "status", "okay");
260         qemu_fdt_setprop_cell(ms->fdt, cpu_name, "reg",
261             s->soc[socket].hartid_base + cpu);
262         qemu_fdt_setprop_string(ms->fdt, cpu_name, "device_type", "cpu");
263         riscv_socket_fdt_write_id(ms, cpu_name, socket);
264         qemu_fdt_setprop_cell(ms->fdt, cpu_name, "phandle", cpu_phandle);
265 
266         intc_phandles[cpu] = (*phandle)++;
267 
268         intc_name = g_strdup_printf("%s/interrupt-controller", cpu_name);
269         qemu_fdt_add_subnode(ms->fdt, intc_name);
270         qemu_fdt_setprop_cell(ms->fdt, intc_name, "phandle",
271             intc_phandles[cpu]);
272         qemu_fdt_setprop_string(ms->fdt, intc_name, "compatible",
273             "riscv,cpu-intc");
274         qemu_fdt_setprop(ms->fdt, intc_name, "interrupt-controller", NULL, 0);
275         qemu_fdt_setprop_cell(ms->fdt, intc_name, "#interrupt-cells", 1);
276 
277         core_name = g_strdup_printf("%s/core%d", clust_name, cpu);
278         qemu_fdt_add_subnode(ms->fdt, core_name);
279         qemu_fdt_setprop_cell(ms->fdt, core_name, "cpu", cpu_phandle);
280 
281         g_free(core_name);
282         g_free(intc_name);
283         g_free(cpu_name);
284     }
285 }
286 
287 static void create_fdt_socket_memory(RISCVVirtState *s,
288                                      const MemMapEntry *memmap, int socket)
289 {
290     char *mem_name;
291     uint64_t addr, size;
292     MachineState *ms = MACHINE(s);
293 
294     addr = memmap[VIRT_DRAM].base + riscv_socket_mem_offset(ms, socket);
295     size = riscv_socket_mem_size(ms, socket);
296     mem_name = g_strdup_printf("/memory@%lx", (long)addr);
297     qemu_fdt_add_subnode(ms->fdt, mem_name);
298     qemu_fdt_setprop_cells(ms->fdt, mem_name, "reg",
299         addr >> 32, addr, size >> 32, size);
300     qemu_fdt_setprop_string(ms->fdt, mem_name, "device_type", "memory");
301     riscv_socket_fdt_write_id(ms, mem_name, socket);
302     g_free(mem_name);
303 }
304 
305 static void create_fdt_socket_clint(RISCVVirtState *s,
306                                     const MemMapEntry *memmap, int socket,
307                                     uint32_t *intc_phandles)
308 {
309     int cpu;
310     char *clint_name;
311     uint32_t *clint_cells;
312     unsigned long clint_addr;
313     MachineState *ms = MACHINE(s);
314     static const char * const clint_compat[2] = {
315         "sifive,clint0", "riscv,clint0"
316     };
317 
318     clint_cells = g_new0(uint32_t, s->soc[socket].num_harts * 4);
319 
320     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
321         clint_cells[cpu * 4 + 0] = cpu_to_be32(intc_phandles[cpu]);
322         clint_cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_SOFT);
323         clint_cells[cpu * 4 + 2] = cpu_to_be32(intc_phandles[cpu]);
324         clint_cells[cpu * 4 + 3] = cpu_to_be32(IRQ_M_TIMER);
325     }
326 
327     clint_addr = memmap[VIRT_CLINT].base + (memmap[VIRT_CLINT].size * socket);
328     clint_name = g_strdup_printf("/soc/clint@%lx", clint_addr);
329     qemu_fdt_add_subnode(ms->fdt, clint_name);
330     qemu_fdt_setprop_string_array(ms->fdt, clint_name, "compatible",
331                                   (char **)&clint_compat,
332                                   ARRAY_SIZE(clint_compat));
333     qemu_fdt_setprop_cells(ms->fdt, clint_name, "reg",
334         0x0, clint_addr, 0x0, memmap[VIRT_CLINT].size);
335     qemu_fdt_setprop(ms->fdt, clint_name, "interrupts-extended",
336         clint_cells, s->soc[socket].num_harts * sizeof(uint32_t) * 4);
337     riscv_socket_fdt_write_id(ms, clint_name, socket);
338     g_free(clint_name);
339 
340     g_free(clint_cells);
341 }
342 
343 static void create_fdt_socket_aclint(RISCVVirtState *s,
344                                      const MemMapEntry *memmap, int socket,
345                                      uint32_t *intc_phandles)
346 {
347     int cpu;
348     char *name;
349     unsigned long addr, size;
350     uint32_t aclint_cells_size;
351     uint32_t *aclint_mswi_cells;
352     uint32_t *aclint_sswi_cells;
353     uint32_t *aclint_mtimer_cells;
354     MachineState *ms = MACHINE(s);
355 
356     aclint_mswi_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
357     aclint_mtimer_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
358     aclint_sswi_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
359 
360     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
361         aclint_mswi_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
362         aclint_mswi_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_M_SOFT);
363         aclint_mtimer_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
364         aclint_mtimer_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_M_TIMER);
365         aclint_sswi_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
366         aclint_sswi_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_S_SOFT);
367     }
368     aclint_cells_size = s->soc[socket].num_harts * sizeof(uint32_t) * 2;
369 
370     if (s->aia_type != VIRT_AIA_TYPE_APLIC_IMSIC) {
371         addr = memmap[VIRT_CLINT].base + (memmap[VIRT_CLINT].size * socket);
372         name = g_strdup_printf("/soc/mswi@%lx", addr);
373         qemu_fdt_add_subnode(ms->fdt, name);
374         qemu_fdt_setprop_string(ms->fdt, name, "compatible",
375             "riscv,aclint-mswi");
376         qemu_fdt_setprop_cells(ms->fdt, name, "reg",
377             0x0, addr, 0x0, RISCV_ACLINT_SWI_SIZE);
378         qemu_fdt_setprop(ms->fdt, name, "interrupts-extended",
379             aclint_mswi_cells, aclint_cells_size);
380         qemu_fdt_setprop(ms->fdt, name, "interrupt-controller", NULL, 0);
381         qemu_fdt_setprop_cell(ms->fdt, name, "#interrupt-cells", 0);
382         riscv_socket_fdt_write_id(ms, name, socket);
383         g_free(name);
384     }
385 
386     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
387         addr = memmap[VIRT_CLINT].base +
388                (RISCV_ACLINT_DEFAULT_MTIMER_SIZE * socket);
389         size = RISCV_ACLINT_DEFAULT_MTIMER_SIZE;
390     } else {
391         addr = memmap[VIRT_CLINT].base + RISCV_ACLINT_SWI_SIZE +
392             (memmap[VIRT_CLINT].size * socket);
393         size = memmap[VIRT_CLINT].size - RISCV_ACLINT_SWI_SIZE;
394     }
395     name = g_strdup_printf("/soc/mtimer@%lx", addr);
396     qemu_fdt_add_subnode(ms->fdt, name);
397     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
398         "riscv,aclint-mtimer");
399     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
400         0x0, addr + RISCV_ACLINT_DEFAULT_MTIME,
401         0x0, size - RISCV_ACLINT_DEFAULT_MTIME,
402         0x0, addr + RISCV_ACLINT_DEFAULT_MTIMECMP,
403         0x0, RISCV_ACLINT_DEFAULT_MTIME);
404     qemu_fdt_setprop(ms->fdt, name, "interrupts-extended",
405         aclint_mtimer_cells, aclint_cells_size);
406     riscv_socket_fdt_write_id(ms, name, socket);
407     g_free(name);
408 
409     if (s->aia_type != VIRT_AIA_TYPE_APLIC_IMSIC) {
410         addr = memmap[VIRT_ACLINT_SSWI].base +
411             (memmap[VIRT_ACLINT_SSWI].size * socket);
412         name = g_strdup_printf("/soc/sswi@%lx", addr);
413         qemu_fdt_add_subnode(ms->fdt, name);
414         qemu_fdt_setprop_string(ms->fdt, name, "compatible",
415             "riscv,aclint-sswi");
416         qemu_fdt_setprop_cells(ms->fdt, name, "reg",
417             0x0, addr, 0x0, memmap[VIRT_ACLINT_SSWI].size);
418         qemu_fdt_setprop(ms->fdt, name, "interrupts-extended",
419             aclint_sswi_cells, aclint_cells_size);
420         qemu_fdt_setprop(ms->fdt, name, "interrupt-controller", NULL, 0);
421         qemu_fdt_setprop_cell(ms->fdt, name, "#interrupt-cells", 0);
422         riscv_socket_fdt_write_id(ms, name, socket);
423         g_free(name);
424     }
425 
426     g_free(aclint_mswi_cells);
427     g_free(aclint_mtimer_cells);
428     g_free(aclint_sswi_cells);
429 }
430 
431 static void create_fdt_socket_plic(RISCVVirtState *s,
432                                    const MemMapEntry *memmap, int socket,
433                                    uint32_t *phandle, uint32_t *intc_phandles,
434                                    uint32_t *plic_phandles)
435 {
436     int cpu;
437     char *plic_name;
438     uint32_t *plic_cells;
439     unsigned long plic_addr;
440     MachineState *ms = MACHINE(s);
441     static const char * const plic_compat[2] = {
442         "sifive,plic-1.0.0", "riscv,plic0"
443     };
444 
445     plic_phandles[socket] = (*phandle)++;
446     plic_addr = memmap[VIRT_PLIC].base + (memmap[VIRT_PLIC].size * socket);
447     plic_name = g_strdup_printf("/soc/plic@%lx", plic_addr);
448     qemu_fdt_add_subnode(ms->fdt, plic_name);
449     qemu_fdt_setprop_cell(ms->fdt, plic_name,
450         "#interrupt-cells", FDT_PLIC_INT_CELLS);
451     qemu_fdt_setprop_cell(ms->fdt, plic_name,
452         "#address-cells", FDT_PLIC_ADDR_CELLS);
453     qemu_fdt_setprop_string_array(ms->fdt, plic_name, "compatible",
454                                   (char **)&plic_compat,
455                                   ARRAY_SIZE(plic_compat));
456     qemu_fdt_setprop(ms->fdt, plic_name, "interrupt-controller", NULL, 0);
457 
458     if (kvm_enabled()) {
459         plic_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
460 
461         for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
462             plic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
463             plic_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_S_EXT);
464         }
465 
466         qemu_fdt_setprop(ms->fdt, plic_name, "interrupts-extended",
467                          plic_cells,
468                          s->soc[socket].num_harts * sizeof(uint32_t) * 2);
469    } else {
470         plic_cells = g_new0(uint32_t, s->soc[socket].num_harts * 4);
471 
472         for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
473             plic_cells[cpu * 4 + 0] = cpu_to_be32(intc_phandles[cpu]);
474             plic_cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_EXT);
475             plic_cells[cpu * 4 + 2] = cpu_to_be32(intc_phandles[cpu]);
476             plic_cells[cpu * 4 + 3] = cpu_to_be32(IRQ_S_EXT);
477         }
478 
479         qemu_fdt_setprop(ms->fdt, plic_name, "interrupts-extended",
480                          plic_cells,
481                          s->soc[socket].num_harts * sizeof(uint32_t) * 4);
482     }
483 
484     qemu_fdt_setprop_cells(ms->fdt, plic_name, "reg",
485         0x0, plic_addr, 0x0, memmap[VIRT_PLIC].size);
486     qemu_fdt_setprop_cell(ms->fdt, plic_name, "riscv,ndev",
487                           VIRT_IRQCHIP_NUM_SOURCES - 1);
488     riscv_socket_fdt_write_id(ms, plic_name, socket);
489     qemu_fdt_setprop_cell(ms->fdt, plic_name, "phandle",
490         plic_phandles[socket]);
491 
492     if (!socket) {
493         platform_bus_add_all_fdt_nodes(ms->fdt, plic_name,
494                                        memmap[VIRT_PLATFORM_BUS].base,
495                                        memmap[VIRT_PLATFORM_BUS].size,
496                                        VIRT_PLATFORM_BUS_IRQ);
497     }
498 
499     g_free(plic_name);
500 
501     g_free(plic_cells);
502 }
503 
504 uint32_t imsic_num_bits(uint32_t count)
505 {
506     uint32_t ret = 0;
507 
508     while (BIT(ret) < count) {
509         ret++;
510     }
511 
512     return ret;
513 }
514 
515 static void create_fdt_one_imsic(RISCVVirtState *s, hwaddr base_addr,
516                                  uint32_t *intc_phandles, uint32_t msi_phandle,
517                                  bool m_mode, uint32_t imsic_guest_bits)
518 {
519     int cpu, socket;
520     char *imsic_name;
521     MachineState *ms = MACHINE(s);
522     int socket_count = riscv_socket_count(ms);
523     uint32_t imsic_max_hart_per_socket;
524     uint32_t *imsic_cells, *imsic_regs, imsic_addr, imsic_size;
525 
526     imsic_cells = g_new0(uint32_t, ms->smp.cpus * 2);
527     imsic_regs = g_new0(uint32_t, socket_count * 4);
528 
529     for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
530         imsic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
531         imsic_cells[cpu * 2 + 1] = cpu_to_be32(m_mode ? IRQ_M_EXT : IRQ_S_EXT);
532     }
533 
534     imsic_max_hart_per_socket = 0;
535     for (socket = 0; socket < socket_count; socket++) {
536         imsic_addr = base_addr + socket * VIRT_IMSIC_GROUP_MAX_SIZE;
537         imsic_size = IMSIC_HART_SIZE(imsic_guest_bits) *
538                      s->soc[socket].num_harts;
539         imsic_regs[socket * 4 + 0] = 0;
540         imsic_regs[socket * 4 + 1] = cpu_to_be32(imsic_addr);
541         imsic_regs[socket * 4 + 2] = 0;
542         imsic_regs[socket * 4 + 3] = cpu_to_be32(imsic_size);
543         if (imsic_max_hart_per_socket < s->soc[socket].num_harts) {
544             imsic_max_hart_per_socket = s->soc[socket].num_harts;
545         }
546     }
547 
548     imsic_name = g_strdup_printf("/soc/imsics@%lx", (unsigned long)base_addr);
549     qemu_fdt_add_subnode(ms->fdt, imsic_name);
550     qemu_fdt_setprop_string(ms->fdt, imsic_name, "compatible", "riscv,imsics");
551     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "#interrupt-cells",
552                           FDT_IMSIC_INT_CELLS);
553     qemu_fdt_setprop(ms->fdt, imsic_name, "interrupt-controller", NULL, 0);
554     qemu_fdt_setprop(ms->fdt, imsic_name, "msi-controller", NULL, 0);
555     qemu_fdt_setprop(ms->fdt, imsic_name, "interrupts-extended",
556                      imsic_cells, ms->smp.cpus * sizeof(uint32_t) * 2);
557     qemu_fdt_setprop(ms->fdt, imsic_name, "reg", imsic_regs,
558                      socket_count * sizeof(uint32_t) * 4);
559     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,num-ids",
560                      VIRT_IRQCHIP_NUM_MSIS);
561 
562     if (imsic_guest_bits) {
563         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,guest-index-bits",
564                               imsic_guest_bits);
565     }
566 
567     if (socket_count > 1) {
568         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,hart-index-bits",
569                               imsic_num_bits(imsic_max_hart_per_socket));
570         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-bits",
571                               imsic_num_bits(socket_count));
572         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-shift",
573                               IMSIC_MMIO_GROUP_MIN_SHIFT);
574     }
575     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "phandle", msi_phandle);
576 
577     g_free(imsic_name);
578     g_free(imsic_regs);
579     g_free(imsic_cells);
580 }
581 
582 static void create_fdt_imsic(RISCVVirtState *s, const MemMapEntry *memmap,
583                              uint32_t *phandle, uint32_t *intc_phandles,
584                              uint32_t *msi_m_phandle, uint32_t *msi_s_phandle)
585 {
586     *msi_m_phandle = (*phandle)++;
587     *msi_s_phandle = (*phandle)++;
588 
589     if (!kvm_enabled()) {
590         /* M-level IMSIC node */
591         create_fdt_one_imsic(s, memmap[VIRT_IMSIC_M].base, intc_phandles,
592                              *msi_m_phandle, true, 0);
593     }
594 
595     /* S-level IMSIC node */
596     create_fdt_one_imsic(s, memmap[VIRT_IMSIC_S].base, intc_phandles,
597                          *msi_s_phandle, false,
598                          imsic_num_bits(s->aia_guests + 1));
599 
600 }
601 
602 static void create_fdt_one_aplic(RISCVVirtState *s, int socket,
603                                  unsigned long aplic_addr, uint32_t aplic_size,
604                                  uint32_t msi_phandle,
605                                  uint32_t *intc_phandles,
606                                  uint32_t aplic_phandle,
607                                  uint32_t aplic_child_phandle,
608                                  bool m_mode, int num_harts)
609 {
610     int cpu;
611     char *aplic_name;
612     uint32_t *aplic_cells;
613     MachineState *ms = MACHINE(s);
614 
615     aplic_cells = g_new0(uint32_t, num_harts * 2);
616 
617     for (cpu = 0; cpu < num_harts; cpu++) {
618         aplic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
619         aplic_cells[cpu * 2 + 1] = cpu_to_be32(m_mode ? IRQ_M_EXT : IRQ_S_EXT);
620     }
621 
622     aplic_name = g_strdup_printf("/soc/aplic@%lx", aplic_addr);
623     qemu_fdt_add_subnode(ms->fdt, aplic_name);
624     qemu_fdt_setprop_string(ms->fdt, aplic_name, "compatible", "riscv,aplic");
625     qemu_fdt_setprop_cell(ms->fdt, aplic_name,
626                           "#interrupt-cells", FDT_APLIC_INT_CELLS);
627     qemu_fdt_setprop(ms->fdt, aplic_name, "interrupt-controller", NULL, 0);
628 
629     if (s->aia_type == VIRT_AIA_TYPE_APLIC) {
630         qemu_fdt_setprop(ms->fdt, aplic_name, "interrupts-extended",
631                          aplic_cells, num_harts * sizeof(uint32_t) * 2);
632     } else {
633         qemu_fdt_setprop_cell(ms->fdt, aplic_name, "msi-parent", msi_phandle);
634     }
635 
636     qemu_fdt_setprop_cells(ms->fdt, aplic_name, "reg",
637                            0x0, aplic_addr, 0x0, aplic_size);
638     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "riscv,num-sources",
639                           VIRT_IRQCHIP_NUM_SOURCES);
640 
641     if (aplic_child_phandle) {
642         qemu_fdt_setprop_cell(ms->fdt, aplic_name, "riscv,children",
643                               aplic_child_phandle);
644         qemu_fdt_setprop_cells(ms->fdt, aplic_name, "riscv,delegate",
645                                aplic_child_phandle, 0x1,
646                                VIRT_IRQCHIP_NUM_SOURCES);
647     }
648 
649     riscv_socket_fdt_write_id(ms, aplic_name, socket);
650     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "phandle", aplic_phandle);
651 
652     g_free(aplic_name);
653     g_free(aplic_cells);
654 }
655 
656 static void create_fdt_socket_aplic(RISCVVirtState *s,
657                                     const MemMapEntry *memmap, int socket,
658                                     uint32_t msi_m_phandle,
659                                     uint32_t msi_s_phandle,
660                                     uint32_t *phandle,
661                                     uint32_t *intc_phandles,
662                                     uint32_t *aplic_phandles,
663                                     int num_harts)
664 {
665     char *aplic_name;
666     unsigned long aplic_addr;
667     MachineState *ms = MACHINE(s);
668     uint32_t aplic_m_phandle, aplic_s_phandle;
669 
670     aplic_m_phandle = (*phandle)++;
671     aplic_s_phandle = (*phandle)++;
672 
673     if (!kvm_enabled()) {
674         /* M-level APLIC node */
675         aplic_addr = memmap[VIRT_APLIC_M].base +
676                      (memmap[VIRT_APLIC_M].size * socket);
677         create_fdt_one_aplic(s, socket, aplic_addr, memmap[VIRT_APLIC_M].size,
678                              msi_m_phandle, intc_phandles,
679                              aplic_m_phandle, aplic_s_phandle,
680                              true, num_harts);
681     }
682 
683     /* S-level APLIC node */
684     aplic_addr = memmap[VIRT_APLIC_S].base +
685                  (memmap[VIRT_APLIC_S].size * socket);
686     create_fdt_one_aplic(s, socket, aplic_addr, memmap[VIRT_APLIC_S].size,
687                          msi_s_phandle, intc_phandles,
688                          aplic_s_phandle, 0,
689                          false, num_harts);
690 
691     aplic_name = g_strdup_printf("/soc/aplic@%lx", aplic_addr);
692 
693     if (!socket) {
694         platform_bus_add_all_fdt_nodes(ms->fdt, aplic_name,
695                                        memmap[VIRT_PLATFORM_BUS].base,
696                                        memmap[VIRT_PLATFORM_BUS].size,
697                                        VIRT_PLATFORM_BUS_IRQ);
698     }
699 
700     g_free(aplic_name);
701 
702     aplic_phandles[socket] = aplic_s_phandle;
703 }
704 
705 static void create_fdt_pmu(RISCVVirtState *s)
706 {
707     char *pmu_name;
708     MachineState *ms = MACHINE(s);
709     RISCVCPU hart = s->soc[0].harts[0];
710 
711     pmu_name = g_strdup_printf("/pmu");
712     qemu_fdt_add_subnode(ms->fdt, pmu_name);
713     qemu_fdt_setprop_string(ms->fdt, pmu_name, "compatible", "riscv,pmu");
714     riscv_pmu_generate_fdt_node(ms->fdt, hart.pmu_avail_ctrs, pmu_name);
715 
716     g_free(pmu_name);
717 }
718 
719 static void create_fdt_sockets(RISCVVirtState *s, const MemMapEntry *memmap,
720                                uint32_t *phandle,
721                                uint32_t *irq_mmio_phandle,
722                                uint32_t *irq_pcie_phandle,
723                                uint32_t *irq_virtio_phandle,
724                                uint32_t *msi_pcie_phandle)
725 {
726     char *clust_name;
727     int socket, phandle_pos;
728     MachineState *ms = MACHINE(s);
729     uint32_t msi_m_phandle = 0, msi_s_phandle = 0;
730     uint32_t *intc_phandles, xplic_phandles[MAX_NODES];
731     int socket_count = riscv_socket_count(ms);
732 
733     qemu_fdt_add_subnode(ms->fdt, "/cpus");
734     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "timebase-frequency",
735                           RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ);
736     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
737     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", 0x1);
738     qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
739 
740     intc_phandles = g_new0(uint32_t, ms->smp.cpus);
741 
742     phandle_pos = ms->smp.cpus;
743     for (socket = (socket_count - 1); socket >= 0; socket--) {
744         phandle_pos -= s->soc[socket].num_harts;
745 
746         clust_name = g_strdup_printf("/cpus/cpu-map/cluster%d", socket);
747         qemu_fdt_add_subnode(ms->fdt, clust_name);
748 
749         create_fdt_socket_cpus(s, socket, clust_name, phandle,
750                                &intc_phandles[phandle_pos]);
751 
752         create_fdt_socket_memory(s, memmap, socket);
753 
754         g_free(clust_name);
755 
756         if (tcg_enabled()) {
757             if (s->have_aclint) {
758                 create_fdt_socket_aclint(s, memmap, socket,
759                     &intc_phandles[phandle_pos]);
760             } else {
761                 create_fdt_socket_clint(s, memmap, socket,
762                     &intc_phandles[phandle_pos]);
763             }
764         }
765     }
766 
767     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
768         create_fdt_imsic(s, memmap, phandle, intc_phandles,
769             &msi_m_phandle, &msi_s_phandle);
770         *msi_pcie_phandle = msi_s_phandle;
771     }
772 
773     /* KVM AIA only has one APLIC instance */
774     if (kvm_enabled() && virt_use_kvm_aia(s)) {
775         create_fdt_socket_aplic(s, memmap, 0,
776                                 msi_m_phandle, msi_s_phandle, phandle,
777                                 &intc_phandles[0], xplic_phandles,
778                                 ms->smp.cpus);
779     } else {
780         phandle_pos = ms->smp.cpus;
781         for (socket = (socket_count - 1); socket >= 0; socket--) {
782             phandle_pos -= s->soc[socket].num_harts;
783 
784             if (s->aia_type == VIRT_AIA_TYPE_NONE) {
785                 create_fdt_socket_plic(s, memmap, socket, phandle,
786                                        &intc_phandles[phandle_pos],
787                                        xplic_phandles);
788             } else {
789                 create_fdt_socket_aplic(s, memmap, socket,
790                                         msi_m_phandle, msi_s_phandle, phandle,
791                                         &intc_phandles[phandle_pos],
792                                         xplic_phandles,
793                                         s->soc[socket].num_harts);
794             }
795         }
796     }
797 
798     g_free(intc_phandles);
799 
800     if (kvm_enabled() && virt_use_kvm_aia(s)) {
801         *irq_mmio_phandle = xplic_phandles[0];
802         *irq_virtio_phandle = xplic_phandles[0];
803         *irq_pcie_phandle = xplic_phandles[0];
804     } else {
805         for (socket = 0; socket < socket_count; socket++) {
806             if (socket == 0) {
807                 *irq_mmio_phandle = xplic_phandles[socket];
808                 *irq_virtio_phandle = xplic_phandles[socket];
809                 *irq_pcie_phandle = xplic_phandles[socket];
810             }
811             if (socket == 1) {
812                 *irq_virtio_phandle = xplic_phandles[socket];
813                 *irq_pcie_phandle = xplic_phandles[socket];
814             }
815             if (socket == 2) {
816                 *irq_pcie_phandle = xplic_phandles[socket];
817             }
818         }
819     }
820 
821     riscv_socket_fdt_write_distance_matrix(ms);
822 }
823 
824 static void create_fdt_virtio(RISCVVirtState *s, const MemMapEntry *memmap,
825                               uint32_t irq_virtio_phandle)
826 {
827     int i;
828     char *name;
829     MachineState *ms = MACHINE(s);
830 
831     for (i = 0; i < VIRTIO_COUNT; i++) {
832         name = g_strdup_printf("/soc/virtio_mmio@%lx",
833             (long)(memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size));
834         qemu_fdt_add_subnode(ms->fdt, name);
835         qemu_fdt_setprop_string(ms->fdt, name, "compatible", "virtio,mmio");
836         qemu_fdt_setprop_cells(ms->fdt, name, "reg",
837             0x0, memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size,
838             0x0, memmap[VIRT_VIRTIO].size);
839         qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent",
840             irq_virtio_phandle);
841         if (s->aia_type == VIRT_AIA_TYPE_NONE) {
842             qemu_fdt_setprop_cell(ms->fdt, name, "interrupts",
843                                   VIRTIO_IRQ + i);
844         } else {
845             qemu_fdt_setprop_cells(ms->fdt, name, "interrupts",
846                                    VIRTIO_IRQ + i, 0x4);
847         }
848         g_free(name);
849     }
850 }
851 
852 static void create_fdt_pcie(RISCVVirtState *s, const MemMapEntry *memmap,
853                             uint32_t irq_pcie_phandle,
854                             uint32_t msi_pcie_phandle)
855 {
856     char *name;
857     MachineState *ms = MACHINE(s);
858 
859     name = g_strdup_printf("/soc/pci@%lx",
860         (long) memmap[VIRT_PCIE_ECAM].base);
861     qemu_fdt_add_subnode(ms->fdt, name);
862     qemu_fdt_setprop_cell(ms->fdt, name, "#address-cells",
863         FDT_PCI_ADDR_CELLS);
864     qemu_fdt_setprop_cell(ms->fdt, name, "#interrupt-cells",
865         FDT_PCI_INT_CELLS);
866     qemu_fdt_setprop_cell(ms->fdt, name, "#size-cells", 0x2);
867     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
868         "pci-host-ecam-generic");
869     qemu_fdt_setprop_string(ms->fdt, name, "device_type", "pci");
870     qemu_fdt_setprop_cell(ms->fdt, name, "linux,pci-domain", 0);
871     qemu_fdt_setprop_cells(ms->fdt, name, "bus-range", 0,
872         memmap[VIRT_PCIE_ECAM].size / PCIE_MMCFG_SIZE_MIN - 1);
873     qemu_fdt_setprop(ms->fdt, name, "dma-coherent", NULL, 0);
874     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
875         qemu_fdt_setprop_cell(ms->fdt, name, "msi-parent", msi_pcie_phandle);
876     }
877     qemu_fdt_setprop_cells(ms->fdt, name, "reg", 0,
878         memmap[VIRT_PCIE_ECAM].base, 0, memmap[VIRT_PCIE_ECAM].size);
879     qemu_fdt_setprop_sized_cells(ms->fdt, name, "ranges",
880         1, FDT_PCI_RANGE_IOPORT, 2, 0,
881         2, memmap[VIRT_PCIE_PIO].base, 2, memmap[VIRT_PCIE_PIO].size,
882         1, FDT_PCI_RANGE_MMIO,
883         2, memmap[VIRT_PCIE_MMIO].base,
884         2, memmap[VIRT_PCIE_MMIO].base, 2, memmap[VIRT_PCIE_MMIO].size,
885         1, FDT_PCI_RANGE_MMIO_64BIT,
886         2, virt_high_pcie_memmap.base,
887         2, virt_high_pcie_memmap.base, 2, virt_high_pcie_memmap.size);
888 
889     create_pcie_irq_map(s, ms->fdt, name, irq_pcie_phandle);
890     g_free(name);
891 }
892 
893 static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap,
894                              uint32_t *phandle)
895 {
896     char *name;
897     uint32_t test_phandle;
898     MachineState *ms = MACHINE(s);
899 
900     test_phandle = (*phandle)++;
901     name = g_strdup_printf("/soc/test@%lx",
902         (long)memmap[VIRT_TEST].base);
903     qemu_fdt_add_subnode(ms->fdt, name);
904     {
905         static const char * const compat[3] = {
906             "sifive,test1", "sifive,test0", "syscon"
907         };
908         qemu_fdt_setprop_string_array(ms->fdt, name, "compatible",
909                                       (char **)&compat, ARRAY_SIZE(compat));
910     }
911     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
912         0x0, memmap[VIRT_TEST].base, 0x0, memmap[VIRT_TEST].size);
913     qemu_fdt_setprop_cell(ms->fdt, name, "phandle", test_phandle);
914     test_phandle = qemu_fdt_get_phandle(ms->fdt, name);
915     g_free(name);
916 
917     name = g_strdup_printf("/reboot");
918     qemu_fdt_add_subnode(ms->fdt, name);
919     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "syscon-reboot");
920     qemu_fdt_setprop_cell(ms->fdt, name, "regmap", test_phandle);
921     qemu_fdt_setprop_cell(ms->fdt, name, "offset", 0x0);
922     qemu_fdt_setprop_cell(ms->fdt, name, "value", FINISHER_RESET);
923     g_free(name);
924 
925     name = g_strdup_printf("/poweroff");
926     qemu_fdt_add_subnode(ms->fdt, name);
927     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "syscon-poweroff");
928     qemu_fdt_setprop_cell(ms->fdt, name, "regmap", test_phandle);
929     qemu_fdt_setprop_cell(ms->fdt, name, "offset", 0x0);
930     qemu_fdt_setprop_cell(ms->fdt, name, "value", FINISHER_PASS);
931     g_free(name);
932 }
933 
934 static void create_fdt_uart(RISCVVirtState *s, const MemMapEntry *memmap,
935                             uint32_t irq_mmio_phandle)
936 {
937     char *name;
938     MachineState *ms = MACHINE(s);
939 
940     name = g_strdup_printf("/soc/serial@%lx", (long)memmap[VIRT_UART0].base);
941     qemu_fdt_add_subnode(ms->fdt, name);
942     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "ns16550a");
943     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
944         0x0, memmap[VIRT_UART0].base,
945         0x0, memmap[VIRT_UART0].size);
946     qemu_fdt_setprop_cell(ms->fdt, name, "clock-frequency", 3686400);
947     qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent", irq_mmio_phandle);
948     if (s->aia_type == VIRT_AIA_TYPE_NONE) {
949         qemu_fdt_setprop_cell(ms->fdt, name, "interrupts", UART0_IRQ);
950     } else {
951         qemu_fdt_setprop_cells(ms->fdt, name, "interrupts", UART0_IRQ, 0x4);
952     }
953 
954     qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", name);
955     g_free(name);
956 }
957 
958 static void create_fdt_rtc(RISCVVirtState *s, const MemMapEntry *memmap,
959                            uint32_t irq_mmio_phandle)
960 {
961     char *name;
962     MachineState *ms = MACHINE(s);
963 
964     name = g_strdup_printf("/soc/rtc@%lx", (long)memmap[VIRT_RTC].base);
965     qemu_fdt_add_subnode(ms->fdt, name);
966     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
967         "google,goldfish-rtc");
968     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
969         0x0, memmap[VIRT_RTC].base, 0x0, memmap[VIRT_RTC].size);
970     qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent",
971         irq_mmio_phandle);
972     if (s->aia_type == VIRT_AIA_TYPE_NONE) {
973         qemu_fdt_setprop_cell(ms->fdt, name, "interrupts", RTC_IRQ);
974     } else {
975         qemu_fdt_setprop_cells(ms->fdt, name, "interrupts", RTC_IRQ, 0x4);
976     }
977     g_free(name);
978 }
979 
980 static void create_fdt_flash(RISCVVirtState *s, const MemMapEntry *memmap)
981 {
982     char *name;
983     MachineState *ms = MACHINE(s);
984     hwaddr flashsize = virt_memmap[VIRT_FLASH].size / 2;
985     hwaddr flashbase = virt_memmap[VIRT_FLASH].base;
986 
987     name = g_strdup_printf("/flash@%" PRIx64, flashbase);
988     qemu_fdt_add_subnode(ms->fdt, name);
989     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "cfi-flash");
990     qemu_fdt_setprop_sized_cells(ms->fdt, name, "reg",
991                                  2, flashbase, 2, flashsize,
992                                  2, flashbase + flashsize, 2, flashsize);
993     qemu_fdt_setprop_cell(ms->fdt, name, "bank-width", 4);
994     g_free(name);
995 }
996 
997 static void create_fdt_fw_cfg(RISCVVirtState *s, const MemMapEntry *memmap)
998 {
999     char *nodename;
1000     MachineState *ms = MACHINE(s);
1001     hwaddr base = memmap[VIRT_FW_CFG].base;
1002     hwaddr size = memmap[VIRT_FW_CFG].size;
1003 
1004     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1005     qemu_fdt_add_subnode(ms->fdt, nodename);
1006     qemu_fdt_setprop_string(ms->fdt, nodename,
1007                             "compatible", "qemu,fw-cfg-mmio");
1008     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1009                                  2, base, 2, size);
1010     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1011     g_free(nodename);
1012 }
1013 
1014 static void finalize_fdt(RISCVVirtState *s)
1015 {
1016     uint32_t phandle = 1, irq_mmio_phandle = 1, msi_pcie_phandle = 1;
1017     uint32_t irq_pcie_phandle = 1, irq_virtio_phandle = 1;
1018 
1019     create_fdt_sockets(s, virt_memmap, &phandle, &irq_mmio_phandle,
1020                        &irq_pcie_phandle, &irq_virtio_phandle,
1021                        &msi_pcie_phandle);
1022 
1023     create_fdt_virtio(s, virt_memmap, irq_virtio_phandle);
1024 
1025     create_fdt_pcie(s, virt_memmap, irq_pcie_phandle, msi_pcie_phandle);
1026 
1027     create_fdt_reset(s, virt_memmap, &phandle);
1028 
1029     create_fdt_uart(s, virt_memmap, irq_mmio_phandle);
1030 
1031     create_fdt_rtc(s, virt_memmap, irq_mmio_phandle);
1032 }
1033 
1034 static void create_fdt(RISCVVirtState *s, const MemMapEntry *memmap)
1035 {
1036     MachineState *ms = MACHINE(s);
1037     uint8_t rng_seed[32];
1038 
1039     ms->fdt = create_device_tree(&s->fdt_size);
1040     if (!ms->fdt) {
1041         error_report("create_device_tree() failed");
1042         exit(1);
1043     }
1044 
1045     qemu_fdt_setprop_string(ms->fdt, "/", "model", "riscv-virtio,qemu");
1046     qemu_fdt_setprop_string(ms->fdt, "/", "compatible", "riscv-virtio");
1047     qemu_fdt_setprop_cell(ms->fdt, "/", "#size-cells", 0x2);
1048     qemu_fdt_setprop_cell(ms->fdt, "/", "#address-cells", 0x2);
1049 
1050     qemu_fdt_add_subnode(ms->fdt, "/soc");
1051     qemu_fdt_setprop(ms->fdt, "/soc", "ranges", NULL, 0);
1052     qemu_fdt_setprop_string(ms->fdt, "/soc", "compatible", "simple-bus");
1053     qemu_fdt_setprop_cell(ms->fdt, "/soc", "#size-cells", 0x2);
1054     qemu_fdt_setprop_cell(ms->fdt, "/soc", "#address-cells", 0x2);
1055 
1056     qemu_fdt_add_subnode(ms->fdt, "/chosen");
1057 
1058     /* Pass seed to RNG */
1059     qemu_guest_getrandom_nofail(rng_seed, sizeof(rng_seed));
1060     qemu_fdt_setprop(ms->fdt, "/chosen", "rng-seed",
1061                      rng_seed, sizeof(rng_seed));
1062 
1063     create_fdt_flash(s, memmap);
1064     create_fdt_fw_cfg(s, memmap);
1065     create_fdt_pmu(s);
1066 }
1067 
1068 static inline DeviceState *gpex_pcie_init(MemoryRegion *sys_mem,
1069                                           DeviceState *irqchip,
1070                                           RISCVVirtState *s)
1071 {
1072     DeviceState *dev;
1073     MemoryRegion *ecam_alias, *ecam_reg;
1074     MemoryRegion *mmio_alias, *high_mmio_alias, *mmio_reg;
1075     hwaddr ecam_base = s->memmap[VIRT_PCIE_ECAM].base;
1076     hwaddr ecam_size = s->memmap[VIRT_PCIE_ECAM].size;
1077     hwaddr mmio_base = s->memmap[VIRT_PCIE_MMIO].base;
1078     hwaddr mmio_size = s->memmap[VIRT_PCIE_MMIO].size;
1079     hwaddr high_mmio_base = virt_high_pcie_memmap.base;
1080     hwaddr high_mmio_size = virt_high_pcie_memmap.size;
1081     hwaddr pio_base = s->memmap[VIRT_PCIE_PIO].base;
1082     hwaddr pio_size = s->memmap[VIRT_PCIE_PIO].size;
1083     qemu_irq irq;
1084     int i;
1085 
1086     dev = qdev_new(TYPE_GPEX_HOST);
1087 
1088     /* Set GPEX object properties for the virt machine */
1089     object_property_set_uint(OBJECT(GPEX_HOST(dev)), PCI_HOST_ECAM_BASE,
1090                             ecam_base, NULL);
1091     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_ECAM_SIZE,
1092                             ecam_size, NULL);
1093     object_property_set_uint(OBJECT(GPEX_HOST(dev)),
1094                              PCI_HOST_BELOW_4G_MMIO_BASE,
1095                              mmio_base, NULL);
1096     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_BELOW_4G_MMIO_SIZE,
1097                             mmio_size, NULL);
1098     object_property_set_uint(OBJECT(GPEX_HOST(dev)),
1099                              PCI_HOST_ABOVE_4G_MMIO_BASE,
1100                              high_mmio_base, NULL);
1101     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_ABOVE_4G_MMIO_SIZE,
1102                             high_mmio_size, NULL);
1103     object_property_set_uint(OBJECT(GPEX_HOST(dev)), PCI_HOST_PIO_BASE,
1104                             pio_base, NULL);
1105     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_PIO_SIZE,
1106                             pio_size, NULL);
1107 
1108     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1109 
1110     ecam_alias = g_new0(MemoryRegion, 1);
1111     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1112     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1113                              ecam_reg, 0, ecam_size);
1114     memory_region_add_subregion(get_system_memory(), ecam_base, ecam_alias);
1115 
1116     mmio_alias = g_new0(MemoryRegion, 1);
1117     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1118     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1119                              mmio_reg, mmio_base, mmio_size);
1120     memory_region_add_subregion(get_system_memory(), mmio_base, mmio_alias);
1121 
1122     /* Map high MMIO space */
1123     high_mmio_alias = g_new0(MemoryRegion, 1);
1124     memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1125                              mmio_reg, high_mmio_base, high_mmio_size);
1126     memory_region_add_subregion(get_system_memory(), high_mmio_base,
1127                                 high_mmio_alias);
1128 
1129     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, pio_base);
1130 
1131     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1132         irq = qdev_get_gpio_in(irqchip, PCIE_IRQ + i);
1133 
1134         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, irq);
1135         gpex_set_irq_num(GPEX_HOST(dev), i, PCIE_IRQ + i);
1136     }
1137 
1138     GPEX_HOST(dev)->gpex_cfg.bus = PCI_HOST_BRIDGE(GPEX_HOST(dev))->bus;
1139     return dev;
1140 }
1141 
1142 static FWCfgState *create_fw_cfg(const MachineState *ms)
1143 {
1144     hwaddr base = virt_memmap[VIRT_FW_CFG].base;
1145     FWCfgState *fw_cfg;
1146 
1147     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16,
1148                                   &address_space_memory);
1149     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1150 
1151     return fw_cfg;
1152 }
1153 
1154 static DeviceState *virt_create_plic(const MemMapEntry *memmap, int socket,
1155                                      int base_hartid, int hart_count)
1156 {
1157     DeviceState *ret;
1158     char *plic_hart_config;
1159 
1160     /* Per-socket PLIC hart topology configuration string */
1161     plic_hart_config = riscv_plic_hart_config_string(hart_count);
1162 
1163     /* Per-socket PLIC */
1164     ret = sifive_plic_create(
1165             memmap[VIRT_PLIC].base + socket * memmap[VIRT_PLIC].size,
1166             plic_hart_config, hart_count, base_hartid,
1167             VIRT_IRQCHIP_NUM_SOURCES,
1168             ((1U << VIRT_IRQCHIP_NUM_PRIO_BITS) - 1),
1169             VIRT_PLIC_PRIORITY_BASE,
1170             VIRT_PLIC_PENDING_BASE,
1171             VIRT_PLIC_ENABLE_BASE,
1172             VIRT_PLIC_ENABLE_STRIDE,
1173             VIRT_PLIC_CONTEXT_BASE,
1174             VIRT_PLIC_CONTEXT_STRIDE,
1175             memmap[VIRT_PLIC].size);
1176 
1177     g_free(plic_hart_config);
1178 
1179     return ret;
1180 }
1181 
1182 static DeviceState *virt_create_aia(RISCVVirtAIAType aia_type, int aia_guests,
1183                                     const MemMapEntry *memmap, int socket,
1184                                     int base_hartid, int hart_count)
1185 {
1186     int i;
1187     hwaddr addr;
1188     uint32_t guest_bits;
1189     DeviceState *aplic_s = NULL;
1190     DeviceState *aplic_m = NULL;
1191     bool msimode = aia_type == VIRT_AIA_TYPE_APLIC_IMSIC;
1192 
1193     if (msimode) {
1194         if (!kvm_enabled()) {
1195             /* Per-socket M-level IMSICs */
1196             addr = memmap[VIRT_IMSIC_M].base +
1197                    socket * VIRT_IMSIC_GROUP_MAX_SIZE;
1198             for (i = 0; i < hart_count; i++) {
1199                 riscv_imsic_create(addr + i * IMSIC_HART_SIZE(0),
1200                                    base_hartid + i, true, 1,
1201                                    VIRT_IRQCHIP_NUM_MSIS);
1202             }
1203         }
1204 
1205         /* Per-socket S-level IMSICs */
1206         guest_bits = imsic_num_bits(aia_guests + 1);
1207         addr = memmap[VIRT_IMSIC_S].base + socket * VIRT_IMSIC_GROUP_MAX_SIZE;
1208         for (i = 0; i < hart_count; i++) {
1209             riscv_imsic_create(addr + i * IMSIC_HART_SIZE(guest_bits),
1210                                base_hartid + i, false, 1 + aia_guests,
1211                                VIRT_IRQCHIP_NUM_MSIS);
1212         }
1213     }
1214 
1215     if (!kvm_enabled()) {
1216         /* Per-socket M-level APLIC */
1217         aplic_m = riscv_aplic_create(memmap[VIRT_APLIC_M].base +
1218                                      socket * memmap[VIRT_APLIC_M].size,
1219                                      memmap[VIRT_APLIC_M].size,
1220                                      (msimode) ? 0 : base_hartid,
1221                                      (msimode) ? 0 : hart_count,
1222                                      VIRT_IRQCHIP_NUM_SOURCES,
1223                                      VIRT_IRQCHIP_NUM_PRIO_BITS,
1224                                      msimode, true, NULL);
1225     }
1226 
1227     /* Per-socket S-level APLIC */
1228     aplic_s = riscv_aplic_create(memmap[VIRT_APLIC_S].base +
1229                                  socket * memmap[VIRT_APLIC_S].size,
1230                                  memmap[VIRT_APLIC_S].size,
1231                                  (msimode) ? 0 : base_hartid,
1232                                  (msimode) ? 0 : hart_count,
1233                                  VIRT_IRQCHIP_NUM_SOURCES,
1234                                  VIRT_IRQCHIP_NUM_PRIO_BITS,
1235                                  msimode, false, aplic_m);
1236 
1237     return kvm_enabled() ? aplic_s : aplic_m;
1238 }
1239 
1240 static void create_platform_bus(RISCVVirtState *s, DeviceState *irqchip)
1241 {
1242     DeviceState *dev;
1243     SysBusDevice *sysbus;
1244     const MemMapEntry *memmap = virt_memmap;
1245     int i;
1246     MemoryRegion *sysmem = get_system_memory();
1247 
1248     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1249     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1250     qdev_prop_set_uint32(dev, "num_irqs", VIRT_PLATFORM_BUS_NUM_IRQS);
1251     qdev_prop_set_uint32(dev, "mmio_size", memmap[VIRT_PLATFORM_BUS].size);
1252     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1253     s->platform_bus_dev = dev;
1254 
1255     sysbus = SYS_BUS_DEVICE(dev);
1256     for (i = 0; i < VIRT_PLATFORM_BUS_NUM_IRQS; i++) {
1257         int irq = VIRT_PLATFORM_BUS_IRQ + i;
1258         sysbus_connect_irq(sysbus, i, qdev_get_gpio_in(irqchip, irq));
1259     }
1260 
1261     memory_region_add_subregion(sysmem,
1262                                 memmap[VIRT_PLATFORM_BUS].base,
1263                                 sysbus_mmio_get_region(sysbus, 0));
1264 }
1265 
1266 static void virt_machine_done(Notifier *notifier, void *data)
1267 {
1268     RISCVVirtState *s = container_of(notifier, RISCVVirtState,
1269                                      machine_done);
1270     const MemMapEntry *memmap = virt_memmap;
1271     MachineState *machine = MACHINE(s);
1272     target_ulong start_addr = memmap[VIRT_DRAM].base;
1273     target_ulong firmware_end_addr, kernel_start_addr;
1274     const char *firmware_name = riscv_default_firmware_name(&s->soc[0]);
1275     uint64_t fdt_load_addr;
1276     uint64_t kernel_entry = 0;
1277     BlockBackend *pflash_blk0;
1278 
1279     /*
1280      * An user provided dtb must include everything, including
1281      * dynamic sysbus devices. Our FDT needs to be finalized.
1282      */
1283     if (machine->dtb == NULL) {
1284         finalize_fdt(s);
1285     }
1286 
1287     /*
1288      * Only direct boot kernel is currently supported for KVM VM,
1289      * so the "-bios" parameter is not supported when KVM is enabled.
1290      */
1291     if (kvm_enabled()) {
1292         if (machine->firmware) {
1293             if (strcmp(machine->firmware, "none")) {
1294                 error_report("Machine mode firmware is not supported in "
1295                              "combination with KVM.");
1296                 exit(1);
1297             }
1298         } else {
1299             machine->firmware = g_strdup("none");
1300         }
1301     }
1302 
1303     firmware_end_addr = riscv_find_and_load_firmware(machine, firmware_name,
1304                                                      start_addr, NULL);
1305 
1306     pflash_blk0 = pflash_cfi01_get_blk(s->flash[0]);
1307     if (pflash_blk0) {
1308         if (machine->firmware && !strcmp(machine->firmware, "none") &&
1309             !kvm_enabled()) {
1310             /*
1311              * Pflash was supplied but bios is none and not KVM guest,
1312              * let's overwrite the address we jump to after reset to
1313              * the base of the flash.
1314              */
1315             start_addr = virt_memmap[VIRT_FLASH].base;
1316         } else {
1317             /*
1318              * Pflash was supplied but either KVM guest or bios is not none.
1319              * In this case, base of the flash would contain S-mode payload.
1320              */
1321             riscv_setup_firmware_boot(machine);
1322             kernel_entry = virt_memmap[VIRT_FLASH].base;
1323         }
1324     }
1325 
1326     if (machine->kernel_filename && !kernel_entry) {
1327         kernel_start_addr = riscv_calc_kernel_start_addr(&s->soc[0],
1328                                                          firmware_end_addr);
1329 
1330         kernel_entry = riscv_load_kernel(machine, &s->soc[0],
1331                                          kernel_start_addr, true, NULL);
1332     }
1333 
1334     fdt_load_addr = riscv_compute_fdt_addr(memmap[VIRT_DRAM].base,
1335                                            memmap[VIRT_DRAM].size,
1336                                            machine);
1337     riscv_load_fdt(fdt_load_addr, machine->fdt);
1338 
1339     /* load the reset vector */
1340     riscv_setup_rom_reset_vec(machine, &s->soc[0], start_addr,
1341                               virt_memmap[VIRT_MROM].base,
1342                               virt_memmap[VIRT_MROM].size, kernel_entry,
1343                               fdt_load_addr);
1344 
1345     /*
1346      * Only direct boot kernel is currently supported for KVM VM,
1347      * So here setup kernel start address and fdt address.
1348      * TODO:Support firmware loading and integrate to TCG start
1349      */
1350     if (kvm_enabled()) {
1351         riscv_setup_direct_kernel(kernel_entry, fdt_load_addr);
1352     }
1353 
1354     if (virt_is_acpi_enabled(s)) {
1355         virt_acpi_setup(s);
1356     }
1357 }
1358 
1359 static void virt_machine_init(MachineState *machine)
1360 {
1361     const MemMapEntry *memmap = virt_memmap;
1362     RISCVVirtState *s = RISCV_VIRT_MACHINE(machine);
1363     MemoryRegion *system_memory = get_system_memory();
1364     MemoryRegion *mask_rom = g_new(MemoryRegion, 1);
1365     char *soc_name;
1366     DeviceState *mmio_irqchip, *virtio_irqchip, *pcie_irqchip;
1367     int i, base_hartid, hart_count;
1368     int socket_count = riscv_socket_count(machine);
1369 
1370     /* Check socket count limit */
1371     if (VIRT_SOCKETS_MAX < socket_count) {
1372         error_report("number of sockets/nodes should be less than %d",
1373             VIRT_SOCKETS_MAX);
1374         exit(1);
1375     }
1376 
1377     if (!tcg_enabled() && s->have_aclint) {
1378         error_report("'aclint' is only available with TCG acceleration");
1379         exit(1);
1380     }
1381 
1382     /* Initialize sockets */
1383     mmio_irqchip = virtio_irqchip = pcie_irqchip = NULL;
1384     for (i = 0; i < socket_count; i++) {
1385         if (!riscv_socket_check_hartids(machine, i)) {
1386             error_report("discontinuous hartids in socket%d", i);
1387             exit(1);
1388         }
1389 
1390         base_hartid = riscv_socket_first_hartid(machine, i);
1391         if (base_hartid < 0) {
1392             error_report("can't find hartid base for socket%d", i);
1393             exit(1);
1394         }
1395 
1396         hart_count = riscv_socket_hart_count(machine, i);
1397         if (hart_count < 0) {
1398             error_report("can't find hart count for socket%d", i);
1399             exit(1);
1400         }
1401 
1402         soc_name = g_strdup_printf("soc%d", i);
1403         object_initialize_child(OBJECT(machine), soc_name, &s->soc[i],
1404                                 TYPE_RISCV_HART_ARRAY);
1405         g_free(soc_name);
1406         object_property_set_str(OBJECT(&s->soc[i]), "cpu-type",
1407                                 machine->cpu_type, &error_abort);
1408         object_property_set_int(OBJECT(&s->soc[i]), "hartid-base",
1409                                 base_hartid, &error_abort);
1410         object_property_set_int(OBJECT(&s->soc[i]), "num-harts",
1411                                 hart_count, &error_abort);
1412         sysbus_realize(SYS_BUS_DEVICE(&s->soc[i]), &error_fatal);
1413 
1414         if (tcg_enabled()) {
1415             if (s->have_aclint) {
1416                 if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
1417                     /* Per-socket ACLINT MTIMER */
1418                     riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1419                             i * RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1420                         RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1421                         base_hartid, hart_count,
1422                         RISCV_ACLINT_DEFAULT_MTIMECMP,
1423                         RISCV_ACLINT_DEFAULT_MTIME,
1424                         RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1425                 } else {
1426                     /* Per-socket ACLINT MSWI, MTIMER, and SSWI */
1427                     riscv_aclint_swi_create(memmap[VIRT_CLINT].base +
1428                             i * memmap[VIRT_CLINT].size,
1429                         base_hartid, hart_count, false);
1430                     riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1431                             i * memmap[VIRT_CLINT].size +
1432                             RISCV_ACLINT_SWI_SIZE,
1433                         RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1434                         base_hartid, hart_count,
1435                         RISCV_ACLINT_DEFAULT_MTIMECMP,
1436                         RISCV_ACLINT_DEFAULT_MTIME,
1437                         RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1438                     riscv_aclint_swi_create(memmap[VIRT_ACLINT_SSWI].base +
1439                             i * memmap[VIRT_ACLINT_SSWI].size,
1440                         base_hartid, hart_count, true);
1441                 }
1442             } else {
1443                 /* Per-socket SiFive CLINT */
1444                 riscv_aclint_swi_create(
1445                     memmap[VIRT_CLINT].base + i * memmap[VIRT_CLINT].size,
1446                     base_hartid, hart_count, false);
1447                 riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1448                         i * memmap[VIRT_CLINT].size + RISCV_ACLINT_SWI_SIZE,
1449                     RISCV_ACLINT_DEFAULT_MTIMER_SIZE, base_hartid, hart_count,
1450                     RISCV_ACLINT_DEFAULT_MTIMECMP, RISCV_ACLINT_DEFAULT_MTIME,
1451                     RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1452             }
1453         }
1454 
1455         /* Per-socket interrupt controller */
1456         if (s->aia_type == VIRT_AIA_TYPE_NONE) {
1457             s->irqchip[i] = virt_create_plic(memmap, i,
1458                                              base_hartid, hart_count);
1459         } else {
1460             s->irqchip[i] = virt_create_aia(s->aia_type, s->aia_guests,
1461                                             memmap, i, base_hartid,
1462                                             hart_count);
1463         }
1464 
1465         /* Try to use different IRQCHIP instance based device type */
1466         if (i == 0) {
1467             mmio_irqchip = s->irqchip[i];
1468             virtio_irqchip = s->irqchip[i];
1469             pcie_irqchip = s->irqchip[i];
1470         }
1471         if (i == 1) {
1472             virtio_irqchip = s->irqchip[i];
1473             pcie_irqchip = s->irqchip[i];
1474         }
1475         if (i == 2) {
1476             pcie_irqchip = s->irqchip[i];
1477         }
1478     }
1479 
1480     if (kvm_enabled() && virt_use_kvm_aia(s)) {
1481         kvm_riscv_aia_create(machine, IMSIC_MMIO_GROUP_MIN_SHIFT,
1482                              VIRT_IRQCHIP_NUM_SOURCES, VIRT_IRQCHIP_NUM_MSIS,
1483                              memmap[VIRT_APLIC_S].base,
1484                              memmap[VIRT_IMSIC_S].base,
1485                              s->aia_guests);
1486     }
1487 
1488     if (riscv_is_32bit(&s->soc[0])) {
1489 #if HOST_LONG_BITS == 64
1490         /* limit RAM size in a 32-bit system */
1491         if (machine->ram_size > 10 * GiB) {
1492             machine->ram_size = 10 * GiB;
1493             error_report("Limiting RAM size to 10 GiB");
1494         }
1495 #endif
1496         virt_high_pcie_memmap.base = VIRT32_HIGH_PCIE_MMIO_BASE;
1497         virt_high_pcie_memmap.size = VIRT32_HIGH_PCIE_MMIO_SIZE;
1498     } else {
1499         virt_high_pcie_memmap.size = VIRT64_HIGH_PCIE_MMIO_SIZE;
1500         virt_high_pcie_memmap.base = memmap[VIRT_DRAM].base + machine->ram_size;
1501         virt_high_pcie_memmap.base =
1502             ROUND_UP(virt_high_pcie_memmap.base, virt_high_pcie_memmap.size);
1503     }
1504 
1505     s->memmap = virt_memmap;
1506 
1507     /* register system main memory (actual RAM) */
1508     memory_region_add_subregion(system_memory, memmap[VIRT_DRAM].base,
1509         machine->ram);
1510 
1511     /* boot rom */
1512     memory_region_init_rom(mask_rom, NULL, "riscv_virt_board.mrom",
1513                            memmap[VIRT_MROM].size, &error_fatal);
1514     memory_region_add_subregion(system_memory, memmap[VIRT_MROM].base,
1515                                 mask_rom);
1516 
1517     /*
1518      * Init fw_cfg. Must be done before riscv_load_fdt, otherwise the
1519      * device tree cannot be altered and we get FDT_ERR_NOSPACE.
1520      */
1521     s->fw_cfg = create_fw_cfg(machine);
1522     rom_set_fw(s->fw_cfg);
1523 
1524     /* SiFive Test MMIO device */
1525     sifive_test_create(memmap[VIRT_TEST].base);
1526 
1527     /* VirtIO MMIO devices */
1528     for (i = 0; i < VIRTIO_COUNT; i++) {
1529         sysbus_create_simple("virtio-mmio",
1530             memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size,
1531             qdev_get_gpio_in(virtio_irqchip, VIRTIO_IRQ + i));
1532     }
1533 
1534     gpex_pcie_init(system_memory, pcie_irqchip, s);
1535 
1536     create_platform_bus(s, mmio_irqchip);
1537 
1538     serial_mm_init(system_memory, memmap[VIRT_UART0].base,
1539         0, qdev_get_gpio_in(mmio_irqchip, UART0_IRQ), 399193,
1540         serial_hd(0), DEVICE_LITTLE_ENDIAN);
1541 
1542     sysbus_create_simple("goldfish_rtc", memmap[VIRT_RTC].base,
1543         qdev_get_gpio_in(mmio_irqchip, RTC_IRQ));
1544 
1545     for (i = 0; i < ARRAY_SIZE(s->flash); i++) {
1546         /* Map legacy -drive if=pflash to machine properties */
1547         pflash_cfi01_legacy_drive(s->flash[i],
1548                                   drive_get(IF_PFLASH, 0, i));
1549     }
1550     virt_flash_map(s, system_memory);
1551 
1552     /* load/create device tree */
1553     if (machine->dtb) {
1554         machine->fdt = load_device_tree(machine->dtb, &s->fdt_size);
1555         if (!machine->fdt) {
1556             error_report("load_device_tree() failed");
1557             exit(1);
1558         }
1559     } else {
1560         create_fdt(s, memmap);
1561     }
1562 
1563     s->machine_done.notify = virt_machine_done;
1564     qemu_add_machine_init_done_notifier(&s->machine_done);
1565 }
1566 
1567 static void virt_machine_instance_init(Object *obj)
1568 {
1569     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1570 
1571     virt_flash_create(s);
1572 
1573     s->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
1574     s->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
1575     s->acpi = ON_OFF_AUTO_AUTO;
1576 }
1577 
1578 static char *virt_get_aia_guests(Object *obj, Error **errp)
1579 {
1580     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1581     char val[32];
1582 
1583     sprintf(val, "%d", s->aia_guests);
1584     return g_strdup(val);
1585 }
1586 
1587 static void virt_set_aia_guests(Object *obj, const char *val, Error **errp)
1588 {
1589     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1590 
1591     s->aia_guests = atoi(val);
1592     if (s->aia_guests < 0 || s->aia_guests > VIRT_IRQCHIP_MAX_GUESTS) {
1593         error_setg(errp, "Invalid number of AIA IMSIC guests");
1594         error_append_hint(errp, "Valid values be between 0 and %d.\n",
1595                           VIRT_IRQCHIP_MAX_GUESTS);
1596     }
1597 }
1598 
1599 static char *virt_get_aia(Object *obj, Error **errp)
1600 {
1601     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1602     const char *val;
1603 
1604     switch (s->aia_type) {
1605     case VIRT_AIA_TYPE_APLIC:
1606         val = "aplic";
1607         break;
1608     case VIRT_AIA_TYPE_APLIC_IMSIC:
1609         val = "aplic-imsic";
1610         break;
1611     default:
1612         val = "none";
1613         break;
1614     };
1615 
1616     return g_strdup(val);
1617 }
1618 
1619 static void virt_set_aia(Object *obj, const char *val, Error **errp)
1620 {
1621     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1622 
1623     if (!strcmp(val, "none")) {
1624         s->aia_type = VIRT_AIA_TYPE_NONE;
1625     } else if (!strcmp(val, "aplic")) {
1626         s->aia_type = VIRT_AIA_TYPE_APLIC;
1627     } else if (!strcmp(val, "aplic-imsic")) {
1628         s->aia_type = VIRT_AIA_TYPE_APLIC_IMSIC;
1629     } else {
1630         error_setg(errp, "Invalid AIA interrupt controller type");
1631         error_append_hint(errp, "Valid values are none, aplic, and "
1632                           "aplic-imsic.\n");
1633     }
1634 }
1635 
1636 static bool virt_get_aclint(Object *obj, Error **errp)
1637 {
1638     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1639 
1640     return s->have_aclint;
1641 }
1642 
1643 static void virt_set_aclint(Object *obj, bool value, Error **errp)
1644 {
1645     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1646 
1647     s->have_aclint = value;
1648 }
1649 
1650 bool virt_is_acpi_enabled(RISCVVirtState *s)
1651 {
1652     return s->acpi != ON_OFF_AUTO_OFF;
1653 }
1654 
1655 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
1656                           void *opaque, Error **errp)
1657 {
1658     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1659     OnOffAuto acpi = s->acpi;
1660 
1661     visit_type_OnOffAuto(v, name, &acpi, errp);
1662 }
1663 
1664 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
1665                           void *opaque, Error **errp)
1666 {
1667     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1668 
1669     visit_type_OnOffAuto(v, name, &s->acpi, errp);
1670 }
1671 
1672 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
1673                                                         DeviceState *dev)
1674 {
1675     MachineClass *mc = MACHINE_GET_CLASS(machine);
1676 
1677     if (device_is_dynamic_sysbus(mc, dev)) {
1678         return HOTPLUG_HANDLER(machine);
1679     }
1680     return NULL;
1681 }
1682 
1683 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1684                                         DeviceState *dev, Error **errp)
1685 {
1686     RISCVVirtState *s = RISCV_VIRT_MACHINE(hotplug_dev);
1687 
1688     if (s->platform_bus_dev) {
1689         MachineClass *mc = MACHINE_GET_CLASS(s);
1690 
1691         if (device_is_dynamic_sysbus(mc, dev)) {
1692             platform_bus_link_device(PLATFORM_BUS_DEVICE(s->platform_bus_dev),
1693                                      SYS_BUS_DEVICE(dev));
1694         }
1695     }
1696 }
1697 
1698 static void virt_machine_class_init(ObjectClass *oc, void *data)
1699 {
1700     char str[128];
1701     MachineClass *mc = MACHINE_CLASS(oc);
1702     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1703 
1704     mc->desc = "RISC-V VirtIO board";
1705     mc->init = virt_machine_init;
1706     mc->max_cpus = VIRT_CPUS_MAX;
1707     mc->default_cpu_type = TYPE_RISCV_CPU_BASE;
1708     mc->pci_allow_0_address = true;
1709     mc->possible_cpu_arch_ids = riscv_numa_possible_cpu_arch_ids;
1710     mc->cpu_index_to_instance_props = riscv_numa_cpu_index_to_props;
1711     mc->get_default_cpu_node_id = riscv_numa_get_default_cpu_node_id;
1712     mc->numa_mem_supported = true;
1713     /* platform instead of architectural choice */
1714     mc->cpu_cluster_has_numa_boundary = true;
1715     mc->default_ram_id = "riscv_virt_board.ram";
1716     assert(!mc->get_hotplug_handler);
1717     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
1718 
1719     hc->plug = virt_machine_device_plug_cb;
1720 
1721     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
1722 #ifdef CONFIG_TPM
1723     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
1724 #endif
1725 
1726 
1727     object_class_property_add_bool(oc, "aclint", virt_get_aclint,
1728                                    virt_set_aclint);
1729     object_class_property_set_description(oc, "aclint",
1730                                           "(TCG only) Set on/off to "
1731                                           "enable/disable emulating "
1732                                           "ACLINT devices");
1733 
1734     object_class_property_add_str(oc, "aia", virt_get_aia,
1735                                   virt_set_aia);
1736     object_class_property_set_description(oc, "aia",
1737                                           "Set type of AIA interrupt "
1738                                           "controller. Valid values are "
1739                                           "none, aplic, and aplic-imsic.");
1740 
1741     object_class_property_add_str(oc, "aia-guests",
1742                                   virt_get_aia_guests,
1743                                   virt_set_aia_guests);
1744     sprintf(str, "Set number of guest MMIO pages for AIA IMSIC. Valid value "
1745                  "should be between 0 and %d.", VIRT_IRQCHIP_MAX_GUESTS);
1746     object_class_property_set_description(oc, "aia-guests", str);
1747     object_class_property_add(oc, "acpi", "OnOffAuto",
1748                               virt_get_acpi, virt_set_acpi,
1749                               NULL, NULL);
1750     object_class_property_set_description(oc, "acpi",
1751                                           "Enable ACPI");
1752 }
1753 
1754 static const TypeInfo virt_machine_typeinfo = {
1755     .name       = MACHINE_TYPE_NAME("virt"),
1756     .parent     = TYPE_MACHINE,
1757     .class_init = virt_machine_class_init,
1758     .instance_init = virt_machine_instance_init,
1759     .instance_size = sizeof(RISCVVirtState),
1760     .interfaces = (InterfaceInfo[]) {
1761          { TYPE_HOTPLUG_HANDLER },
1762          { }
1763     },
1764 };
1765 
1766 static void virt_machine_init_register_types(void)
1767 {
1768     type_register_static(&virt_machine_typeinfo);
1769 }
1770 
1771 type_init(virt_machine_init_register_types)
1772