xref: /openbmc/qemu/hw/riscv/virt.c (revision d2dfe0b5)
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 "hw/intc/riscv_aclint.h"
39 #include "hw/intc/riscv_aplic.h"
40 #include "hw/intc/riscv_imsic.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/kvm.h"
48 #include "sysemu/tpm.h"
49 #include "hw/pci/pci.h"
50 #include "hw/pci-host/gpex.h"
51 #include "hw/display/ramfb.h"
52 #include "hw/acpi/aml-build.h"
53 #include "qapi/qapi-visit-common.h"
54 
55 /*
56  * The virt machine physical address space used by some of the devices
57  * namely ACLINT, PLIC, APLIC, and IMSIC depend on number of Sockets,
58  * number of CPUs, and number of IMSIC guest files.
59  *
60  * Various limits defined by VIRT_SOCKETS_MAX_BITS, VIRT_CPUS_MAX_BITS,
61  * and VIRT_IRQCHIP_MAX_GUESTS_BITS are tuned for maximum utilization
62  * of virt machine physical address space.
63  */
64 
65 #define VIRT_IMSIC_GROUP_MAX_SIZE      (1U << IMSIC_MMIO_GROUP_MIN_SHIFT)
66 #if VIRT_IMSIC_GROUP_MAX_SIZE < \
67     IMSIC_GROUP_SIZE(VIRT_CPUS_MAX_BITS, VIRT_IRQCHIP_MAX_GUESTS_BITS)
68 #error "Can't accomodate single IMSIC group in address space"
69 #endif
70 
71 #define VIRT_IMSIC_MAX_SIZE            (VIRT_SOCKETS_MAX * \
72                                         VIRT_IMSIC_GROUP_MAX_SIZE)
73 #if 0x4000000 < VIRT_IMSIC_MAX_SIZE
74 #error "Can't accomodate all IMSIC groups in address space"
75 #endif
76 
77 static const MemMapEntry virt_memmap[] = {
78     [VIRT_DEBUG] =        {        0x0,         0x100 },
79     [VIRT_MROM] =         {     0x1000,        0xf000 },
80     [VIRT_TEST] =         {   0x100000,        0x1000 },
81     [VIRT_RTC] =          {   0x101000,        0x1000 },
82     [VIRT_CLINT] =        {  0x2000000,       0x10000 },
83     [VIRT_ACLINT_SSWI] =  {  0x2F00000,        0x4000 },
84     [VIRT_PCIE_PIO] =     {  0x3000000,       0x10000 },
85     [VIRT_PLATFORM_BUS] = {  0x4000000,     0x2000000 },
86     [VIRT_PLIC] =         {  0xc000000, VIRT_PLIC_SIZE(VIRT_CPUS_MAX * 2) },
87     [VIRT_APLIC_M] =      {  0xc000000, APLIC_SIZE(VIRT_CPUS_MAX) },
88     [VIRT_APLIC_S] =      {  0xd000000, APLIC_SIZE(VIRT_CPUS_MAX) },
89     [VIRT_UART0] =        { 0x10000000,         0x100 },
90     [VIRT_VIRTIO] =       { 0x10001000,        0x1000 },
91     [VIRT_FW_CFG] =       { 0x10100000,          0x18 },
92     [VIRT_FLASH] =        { 0x20000000,     0x4000000 },
93     [VIRT_IMSIC_M] =      { 0x24000000, VIRT_IMSIC_MAX_SIZE },
94     [VIRT_IMSIC_S] =      { 0x28000000, VIRT_IMSIC_MAX_SIZE },
95     [VIRT_PCIE_ECAM] =    { 0x30000000,    0x10000000 },
96     [VIRT_PCIE_MMIO] =    { 0x40000000,    0x40000000 },
97     [VIRT_DRAM] =         { 0x80000000,           0x0 },
98 };
99 
100 /* PCIe high mmio is fixed for RV32 */
101 #define VIRT32_HIGH_PCIE_MMIO_BASE  0x300000000ULL
102 #define VIRT32_HIGH_PCIE_MMIO_SIZE  (4 * GiB)
103 
104 /* PCIe high mmio for RV64, size is fixed but base depends on top of RAM */
105 #define VIRT64_HIGH_PCIE_MMIO_SIZE  (16 * GiB)
106 
107 static MemMapEntry virt_high_pcie_memmap;
108 
109 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
110 
111 static PFlashCFI01 *virt_flash_create1(RISCVVirtState *s,
112                                        const char *name,
113                                        const char *alias_prop_name)
114 {
115     /*
116      * Create a single flash device.  We use the same parameters as
117      * the flash devices on the ARM virt board.
118      */
119     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
120 
121     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
122     qdev_prop_set_uint8(dev, "width", 4);
123     qdev_prop_set_uint8(dev, "device-width", 2);
124     qdev_prop_set_bit(dev, "big-endian", false);
125     qdev_prop_set_uint16(dev, "id0", 0x89);
126     qdev_prop_set_uint16(dev, "id1", 0x18);
127     qdev_prop_set_uint16(dev, "id2", 0x00);
128     qdev_prop_set_uint16(dev, "id3", 0x00);
129     qdev_prop_set_string(dev, "name", name);
130 
131     object_property_add_child(OBJECT(s), name, OBJECT(dev));
132     object_property_add_alias(OBJECT(s), alias_prop_name,
133                               OBJECT(dev), "drive");
134 
135     return PFLASH_CFI01(dev);
136 }
137 
138 static void virt_flash_create(RISCVVirtState *s)
139 {
140     s->flash[0] = virt_flash_create1(s, "virt.flash0", "pflash0");
141     s->flash[1] = virt_flash_create1(s, "virt.flash1", "pflash1");
142 }
143 
144 static void virt_flash_map1(PFlashCFI01 *flash,
145                             hwaddr base, hwaddr size,
146                             MemoryRegion *sysmem)
147 {
148     DeviceState *dev = DEVICE(flash);
149 
150     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
151     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
152     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
153     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
154 
155     memory_region_add_subregion(sysmem, base,
156                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
157                                                        0));
158 }
159 
160 static void virt_flash_map(RISCVVirtState *s,
161                            MemoryRegion *sysmem)
162 {
163     hwaddr flashsize = virt_memmap[VIRT_FLASH].size / 2;
164     hwaddr flashbase = virt_memmap[VIRT_FLASH].base;
165 
166     virt_flash_map1(s->flash[0], flashbase, flashsize,
167                     sysmem);
168     virt_flash_map1(s->flash[1], flashbase + flashsize, flashsize,
169                     sysmem);
170 }
171 
172 static void create_pcie_irq_map(RISCVVirtState *s, void *fdt, char *nodename,
173                                 uint32_t irqchip_phandle)
174 {
175     int pin, dev;
176     uint32_t irq_map_stride = 0;
177     uint32_t full_irq_map[GPEX_NUM_IRQS * GPEX_NUM_IRQS *
178                           FDT_MAX_INT_MAP_WIDTH] = {};
179     uint32_t *irq_map = full_irq_map;
180 
181     /* This code creates a standard swizzle of interrupts such that
182      * each device's first interrupt is based on it's PCI_SLOT number.
183      * (See pci_swizzle_map_irq_fn())
184      *
185      * We only need one entry per interrupt in the table (not one per
186      * possible slot) seeing the interrupt-map-mask will allow the table
187      * to wrap to any number of devices.
188      */
189     for (dev = 0; dev < GPEX_NUM_IRQS; dev++) {
190         int devfn = dev * 0x8;
191 
192         for (pin = 0; pin < GPEX_NUM_IRQS; pin++) {
193             int irq_nr = PCIE_IRQ + ((pin + PCI_SLOT(devfn)) % GPEX_NUM_IRQS);
194             int i = 0;
195 
196             /* Fill PCI address cells */
197             irq_map[i] = cpu_to_be32(devfn << 8);
198             i += FDT_PCI_ADDR_CELLS;
199 
200             /* Fill PCI Interrupt cells */
201             irq_map[i] = cpu_to_be32(pin + 1);
202             i += FDT_PCI_INT_CELLS;
203 
204             /* Fill interrupt controller phandle and cells */
205             irq_map[i++] = cpu_to_be32(irqchip_phandle);
206             irq_map[i++] = cpu_to_be32(irq_nr);
207             if (s->aia_type != VIRT_AIA_TYPE_NONE) {
208                 irq_map[i++] = cpu_to_be32(0x4);
209             }
210 
211             if (!irq_map_stride) {
212                 irq_map_stride = i;
213             }
214             irq_map += irq_map_stride;
215         }
216     }
217 
218     qemu_fdt_setprop(fdt, nodename, "interrupt-map", full_irq_map,
219                      GPEX_NUM_IRQS * GPEX_NUM_IRQS *
220                      irq_map_stride * sizeof(uint32_t));
221 
222     qemu_fdt_setprop_cells(fdt, nodename, "interrupt-map-mask",
223                            0x1800, 0, 0, 0x7);
224 }
225 
226 static void create_fdt_socket_cpus(RISCVVirtState *s, int socket,
227                                    char *clust_name, uint32_t *phandle,
228                                    uint32_t *intc_phandles)
229 {
230     int cpu;
231     uint32_t cpu_phandle;
232     MachineState *ms = MACHINE(s);
233     char *name, *cpu_name, *core_name, *intc_name, *sv_name;
234     bool is_32_bit = riscv_is_32bit(&s->soc[0]);
235     uint8_t satp_mode_max;
236 
237     for (cpu = s->soc[socket].num_harts - 1; cpu >= 0; cpu--) {
238         RISCVCPU *cpu_ptr = &s->soc[socket].harts[cpu];
239 
240         cpu_phandle = (*phandle)++;
241 
242         cpu_name = g_strdup_printf("/cpus/cpu@%d",
243             s->soc[socket].hartid_base + cpu);
244         qemu_fdt_add_subnode(ms->fdt, cpu_name);
245 
246         satp_mode_max = satp_mode_max_from_map(
247             s->soc[socket].harts[cpu].cfg.satp_mode.map);
248         sv_name = g_strdup_printf("riscv,%s",
249                                   satp_mode_str(satp_mode_max, is_32_bit));
250         qemu_fdt_setprop_string(ms->fdt, cpu_name, "mmu-type", sv_name);
251         g_free(sv_name);
252 
253 
254         name = riscv_isa_string(cpu_ptr);
255         qemu_fdt_setprop_string(ms->fdt, cpu_name, "riscv,isa", name);
256         g_free(name);
257 
258         if (cpu_ptr->cfg.ext_icbom) {
259             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cbom-block-size",
260                                   cpu_ptr->cfg.cbom_blocksize);
261         }
262 
263         if (cpu_ptr->cfg.ext_icboz) {
264             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cboz-block-size",
265                                   cpu_ptr->cfg.cboz_blocksize);
266         }
267 
268         qemu_fdt_setprop_string(ms->fdt, cpu_name, "compatible", "riscv");
269         qemu_fdt_setprop_string(ms->fdt, cpu_name, "status", "okay");
270         qemu_fdt_setprop_cell(ms->fdt, cpu_name, "reg",
271             s->soc[socket].hartid_base + cpu);
272         qemu_fdt_setprop_string(ms->fdt, cpu_name, "device_type", "cpu");
273         riscv_socket_fdt_write_id(ms, cpu_name, socket);
274         qemu_fdt_setprop_cell(ms->fdt, cpu_name, "phandle", cpu_phandle);
275 
276         intc_phandles[cpu] = (*phandle)++;
277 
278         intc_name = g_strdup_printf("%s/interrupt-controller", cpu_name);
279         qemu_fdt_add_subnode(ms->fdt, intc_name);
280         qemu_fdt_setprop_cell(ms->fdt, intc_name, "phandle",
281             intc_phandles[cpu]);
282         qemu_fdt_setprop_string(ms->fdt, intc_name, "compatible",
283             "riscv,cpu-intc");
284         qemu_fdt_setprop(ms->fdt, intc_name, "interrupt-controller", NULL, 0);
285         qemu_fdt_setprop_cell(ms->fdt, intc_name, "#interrupt-cells", 1);
286 
287         core_name = g_strdup_printf("%s/core%d", clust_name, cpu);
288         qemu_fdt_add_subnode(ms->fdt, core_name);
289         qemu_fdt_setprop_cell(ms->fdt, core_name, "cpu", cpu_phandle);
290 
291         g_free(core_name);
292         g_free(intc_name);
293         g_free(cpu_name);
294     }
295 }
296 
297 static void create_fdt_socket_memory(RISCVVirtState *s,
298                                      const MemMapEntry *memmap, int socket)
299 {
300     char *mem_name;
301     uint64_t addr, size;
302     MachineState *ms = MACHINE(s);
303 
304     addr = memmap[VIRT_DRAM].base + riscv_socket_mem_offset(ms, socket);
305     size = riscv_socket_mem_size(ms, socket);
306     mem_name = g_strdup_printf("/memory@%lx", (long)addr);
307     qemu_fdt_add_subnode(ms->fdt, mem_name);
308     qemu_fdt_setprop_cells(ms->fdt, mem_name, "reg",
309         addr >> 32, addr, size >> 32, size);
310     qemu_fdt_setprop_string(ms->fdt, mem_name, "device_type", "memory");
311     riscv_socket_fdt_write_id(ms, mem_name, socket);
312     g_free(mem_name);
313 }
314 
315 static void create_fdt_socket_clint(RISCVVirtState *s,
316                                     const MemMapEntry *memmap, int socket,
317                                     uint32_t *intc_phandles)
318 {
319     int cpu;
320     char *clint_name;
321     uint32_t *clint_cells;
322     unsigned long clint_addr;
323     MachineState *ms = MACHINE(s);
324     static const char * const clint_compat[2] = {
325         "sifive,clint0", "riscv,clint0"
326     };
327 
328     clint_cells = g_new0(uint32_t, s->soc[socket].num_harts * 4);
329 
330     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
331         clint_cells[cpu * 4 + 0] = cpu_to_be32(intc_phandles[cpu]);
332         clint_cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_SOFT);
333         clint_cells[cpu * 4 + 2] = cpu_to_be32(intc_phandles[cpu]);
334         clint_cells[cpu * 4 + 3] = cpu_to_be32(IRQ_M_TIMER);
335     }
336 
337     clint_addr = memmap[VIRT_CLINT].base + (memmap[VIRT_CLINT].size * socket);
338     clint_name = g_strdup_printf("/soc/clint@%lx", clint_addr);
339     qemu_fdt_add_subnode(ms->fdt, clint_name);
340     qemu_fdt_setprop_string_array(ms->fdt, clint_name, "compatible",
341                                   (char **)&clint_compat,
342                                   ARRAY_SIZE(clint_compat));
343     qemu_fdt_setprop_cells(ms->fdt, clint_name, "reg",
344         0x0, clint_addr, 0x0, memmap[VIRT_CLINT].size);
345     qemu_fdt_setprop(ms->fdt, clint_name, "interrupts-extended",
346         clint_cells, s->soc[socket].num_harts * sizeof(uint32_t) * 4);
347     riscv_socket_fdt_write_id(ms, clint_name, socket);
348     g_free(clint_name);
349 
350     g_free(clint_cells);
351 }
352 
353 static void create_fdt_socket_aclint(RISCVVirtState *s,
354                                      const MemMapEntry *memmap, int socket,
355                                      uint32_t *intc_phandles)
356 {
357     int cpu;
358     char *name;
359     unsigned long addr, size;
360     uint32_t aclint_cells_size;
361     uint32_t *aclint_mswi_cells;
362     uint32_t *aclint_sswi_cells;
363     uint32_t *aclint_mtimer_cells;
364     MachineState *ms = MACHINE(s);
365 
366     aclint_mswi_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
367     aclint_mtimer_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
368     aclint_sswi_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
369 
370     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
371         aclint_mswi_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
372         aclint_mswi_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_M_SOFT);
373         aclint_mtimer_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
374         aclint_mtimer_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_M_TIMER);
375         aclint_sswi_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
376         aclint_sswi_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_S_SOFT);
377     }
378     aclint_cells_size = s->soc[socket].num_harts * sizeof(uint32_t) * 2;
379 
380     if (s->aia_type != VIRT_AIA_TYPE_APLIC_IMSIC) {
381         addr = memmap[VIRT_CLINT].base + (memmap[VIRT_CLINT].size * socket);
382         name = g_strdup_printf("/soc/mswi@%lx", addr);
383         qemu_fdt_add_subnode(ms->fdt, name);
384         qemu_fdt_setprop_string(ms->fdt, name, "compatible",
385             "riscv,aclint-mswi");
386         qemu_fdt_setprop_cells(ms->fdt, name, "reg",
387             0x0, addr, 0x0, RISCV_ACLINT_SWI_SIZE);
388         qemu_fdt_setprop(ms->fdt, name, "interrupts-extended",
389             aclint_mswi_cells, aclint_cells_size);
390         qemu_fdt_setprop(ms->fdt, name, "interrupt-controller", NULL, 0);
391         qemu_fdt_setprop_cell(ms->fdt, name, "#interrupt-cells", 0);
392         riscv_socket_fdt_write_id(ms, name, socket);
393         g_free(name);
394     }
395 
396     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
397         addr = memmap[VIRT_CLINT].base +
398                (RISCV_ACLINT_DEFAULT_MTIMER_SIZE * socket);
399         size = RISCV_ACLINT_DEFAULT_MTIMER_SIZE;
400     } else {
401         addr = memmap[VIRT_CLINT].base + RISCV_ACLINT_SWI_SIZE +
402             (memmap[VIRT_CLINT].size * socket);
403         size = memmap[VIRT_CLINT].size - RISCV_ACLINT_SWI_SIZE;
404     }
405     name = g_strdup_printf("/soc/mtimer@%lx", addr);
406     qemu_fdt_add_subnode(ms->fdt, name);
407     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
408         "riscv,aclint-mtimer");
409     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
410         0x0, addr + RISCV_ACLINT_DEFAULT_MTIME,
411         0x0, size - RISCV_ACLINT_DEFAULT_MTIME,
412         0x0, addr + RISCV_ACLINT_DEFAULT_MTIMECMP,
413         0x0, RISCV_ACLINT_DEFAULT_MTIME);
414     qemu_fdt_setprop(ms->fdt, name, "interrupts-extended",
415         aclint_mtimer_cells, aclint_cells_size);
416     riscv_socket_fdt_write_id(ms, name, socket);
417     g_free(name);
418 
419     if (s->aia_type != VIRT_AIA_TYPE_APLIC_IMSIC) {
420         addr = memmap[VIRT_ACLINT_SSWI].base +
421             (memmap[VIRT_ACLINT_SSWI].size * socket);
422         name = g_strdup_printf("/soc/sswi@%lx", addr);
423         qemu_fdt_add_subnode(ms->fdt, name);
424         qemu_fdt_setprop_string(ms->fdt, name, "compatible",
425             "riscv,aclint-sswi");
426         qemu_fdt_setprop_cells(ms->fdt, name, "reg",
427             0x0, addr, 0x0, memmap[VIRT_ACLINT_SSWI].size);
428         qemu_fdt_setprop(ms->fdt, name, "interrupts-extended",
429             aclint_sswi_cells, aclint_cells_size);
430         qemu_fdt_setprop(ms->fdt, name, "interrupt-controller", NULL, 0);
431         qemu_fdt_setprop_cell(ms->fdt, name, "#interrupt-cells", 0);
432         riscv_socket_fdt_write_id(ms, name, socket);
433         g_free(name);
434     }
435 
436     g_free(aclint_mswi_cells);
437     g_free(aclint_mtimer_cells);
438     g_free(aclint_sswi_cells);
439 }
440 
441 static void create_fdt_socket_plic(RISCVVirtState *s,
442                                    const MemMapEntry *memmap, int socket,
443                                    uint32_t *phandle, uint32_t *intc_phandles,
444                                    uint32_t *plic_phandles)
445 {
446     int cpu;
447     char *plic_name;
448     uint32_t *plic_cells;
449     unsigned long plic_addr;
450     MachineState *ms = MACHINE(s);
451     static const char * const plic_compat[2] = {
452         "sifive,plic-1.0.0", "riscv,plic0"
453     };
454 
455     if (kvm_enabled()) {
456         plic_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
457     } else {
458         plic_cells = g_new0(uint32_t, s->soc[socket].num_harts * 4);
459     }
460 
461     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
462         if (kvm_enabled()) {
463             plic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
464             plic_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_S_EXT);
465         } else {
466             plic_cells[cpu * 4 + 0] = cpu_to_be32(intc_phandles[cpu]);
467             plic_cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_EXT);
468             plic_cells[cpu * 4 + 2] = cpu_to_be32(intc_phandles[cpu]);
469             plic_cells[cpu * 4 + 3] = cpu_to_be32(IRQ_S_EXT);
470         }
471     }
472 
473     plic_phandles[socket] = (*phandle)++;
474     plic_addr = memmap[VIRT_PLIC].base + (memmap[VIRT_PLIC].size * socket);
475     plic_name = g_strdup_printf("/soc/plic@%lx", plic_addr);
476     qemu_fdt_add_subnode(ms->fdt, plic_name);
477     qemu_fdt_setprop_cell(ms->fdt, plic_name,
478         "#interrupt-cells", FDT_PLIC_INT_CELLS);
479     qemu_fdt_setprop_cell(ms->fdt, plic_name,
480         "#address-cells", FDT_PLIC_ADDR_CELLS);
481     qemu_fdt_setprop_string_array(ms->fdt, plic_name, "compatible",
482                                   (char **)&plic_compat,
483                                   ARRAY_SIZE(plic_compat));
484     qemu_fdt_setprop(ms->fdt, plic_name, "interrupt-controller", NULL, 0);
485     qemu_fdt_setprop(ms->fdt, plic_name, "interrupts-extended",
486         plic_cells, s->soc[socket].num_harts * sizeof(uint32_t) * 4);
487     qemu_fdt_setprop_cells(ms->fdt, plic_name, "reg",
488         0x0, plic_addr, 0x0, memmap[VIRT_PLIC].size);
489     qemu_fdt_setprop_cell(ms->fdt, plic_name, "riscv,ndev",
490                           VIRT_IRQCHIP_NUM_SOURCES - 1);
491     riscv_socket_fdt_write_id(ms, plic_name, socket);
492     qemu_fdt_setprop_cell(ms->fdt, plic_name, "phandle",
493         plic_phandles[socket]);
494 
495     if (!socket) {
496         platform_bus_add_all_fdt_nodes(ms->fdt, plic_name,
497                                        memmap[VIRT_PLATFORM_BUS].base,
498                                        memmap[VIRT_PLATFORM_BUS].size,
499                                        VIRT_PLATFORM_BUS_IRQ);
500     }
501 
502     g_free(plic_name);
503 
504     g_free(plic_cells);
505 }
506 
507 static uint32_t imsic_num_bits(uint32_t count)
508 {
509     uint32_t ret = 0;
510 
511     while (BIT(ret) < count) {
512         ret++;
513     }
514 
515     return ret;
516 }
517 
518 static void create_fdt_imsic(RISCVVirtState *s, const MemMapEntry *memmap,
519                              uint32_t *phandle, uint32_t *intc_phandles,
520                              uint32_t *msi_m_phandle, uint32_t *msi_s_phandle)
521 {
522     int cpu, socket;
523     char *imsic_name;
524     MachineState *ms = MACHINE(s);
525     int socket_count = riscv_socket_count(ms);
526     uint32_t imsic_max_hart_per_socket, imsic_guest_bits;
527     uint32_t *imsic_cells, *imsic_regs, imsic_addr, imsic_size;
528 
529     *msi_m_phandle = (*phandle)++;
530     *msi_s_phandle = (*phandle)++;
531     imsic_cells = g_new0(uint32_t, ms->smp.cpus * 2);
532     imsic_regs = g_new0(uint32_t, socket_count * 4);
533 
534     /* M-level IMSIC node */
535     for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
536         imsic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
537         imsic_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_M_EXT);
538     }
539     imsic_max_hart_per_socket = 0;
540     for (socket = 0; socket < socket_count; socket++) {
541         imsic_addr = memmap[VIRT_IMSIC_M].base +
542                      socket * VIRT_IMSIC_GROUP_MAX_SIZE;
543         imsic_size = IMSIC_HART_SIZE(0) * s->soc[socket].num_harts;
544         imsic_regs[socket * 4 + 0] = 0;
545         imsic_regs[socket * 4 + 1] = cpu_to_be32(imsic_addr);
546         imsic_regs[socket * 4 + 2] = 0;
547         imsic_regs[socket * 4 + 3] = cpu_to_be32(imsic_size);
548         if (imsic_max_hart_per_socket < s->soc[socket].num_harts) {
549             imsic_max_hart_per_socket = s->soc[socket].num_harts;
550         }
551     }
552     imsic_name = g_strdup_printf("/soc/imsics@%lx",
553         (unsigned long)memmap[VIRT_IMSIC_M].base);
554     qemu_fdt_add_subnode(ms->fdt, imsic_name);
555     qemu_fdt_setprop_string(ms->fdt, imsic_name, "compatible",
556         "riscv,imsics");
557     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "#interrupt-cells",
558         FDT_IMSIC_INT_CELLS);
559     qemu_fdt_setprop(ms->fdt, imsic_name, "interrupt-controller",
560         NULL, 0);
561     qemu_fdt_setprop(ms->fdt, imsic_name, "msi-controller",
562         NULL, 0);
563     qemu_fdt_setprop(ms->fdt, imsic_name, "interrupts-extended",
564         imsic_cells, ms->smp.cpus * sizeof(uint32_t) * 2);
565     qemu_fdt_setprop(ms->fdt, imsic_name, "reg", imsic_regs,
566         socket_count * sizeof(uint32_t) * 4);
567     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,num-ids",
568         VIRT_IRQCHIP_NUM_MSIS);
569     if (socket_count > 1) {
570         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,hart-index-bits",
571             imsic_num_bits(imsic_max_hart_per_socket));
572         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-bits",
573             imsic_num_bits(socket_count));
574         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-shift",
575             IMSIC_MMIO_GROUP_MIN_SHIFT);
576     }
577     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "phandle", *msi_m_phandle);
578 
579     g_free(imsic_name);
580 
581     /* S-level IMSIC node */
582     for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
583         imsic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
584         imsic_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_S_EXT);
585     }
586     imsic_guest_bits = imsic_num_bits(s->aia_guests + 1);
587     imsic_max_hart_per_socket = 0;
588     for (socket = 0; socket < socket_count; socket++) {
589         imsic_addr = memmap[VIRT_IMSIC_S].base +
590                      socket * VIRT_IMSIC_GROUP_MAX_SIZE;
591         imsic_size = IMSIC_HART_SIZE(imsic_guest_bits) *
592                      s->soc[socket].num_harts;
593         imsic_regs[socket * 4 + 0] = 0;
594         imsic_regs[socket * 4 + 1] = cpu_to_be32(imsic_addr);
595         imsic_regs[socket * 4 + 2] = 0;
596         imsic_regs[socket * 4 + 3] = cpu_to_be32(imsic_size);
597         if (imsic_max_hart_per_socket < s->soc[socket].num_harts) {
598             imsic_max_hart_per_socket = s->soc[socket].num_harts;
599         }
600     }
601     imsic_name = g_strdup_printf("/soc/imsics@%lx",
602         (unsigned long)memmap[VIRT_IMSIC_S].base);
603     qemu_fdt_add_subnode(ms->fdt, imsic_name);
604     qemu_fdt_setprop_string(ms->fdt, imsic_name, "compatible",
605         "riscv,imsics");
606     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "#interrupt-cells",
607         FDT_IMSIC_INT_CELLS);
608     qemu_fdt_setprop(ms->fdt, imsic_name, "interrupt-controller",
609         NULL, 0);
610     qemu_fdt_setprop(ms->fdt, imsic_name, "msi-controller",
611         NULL, 0);
612     qemu_fdt_setprop(ms->fdt, imsic_name, "interrupts-extended",
613         imsic_cells, ms->smp.cpus * sizeof(uint32_t) * 2);
614     qemu_fdt_setprop(ms->fdt, imsic_name, "reg", imsic_regs,
615         socket_count * sizeof(uint32_t) * 4);
616     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,num-ids",
617         VIRT_IRQCHIP_NUM_MSIS);
618     if (imsic_guest_bits) {
619         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,guest-index-bits",
620             imsic_guest_bits);
621     }
622     if (socket_count > 1) {
623         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,hart-index-bits",
624             imsic_num_bits(imsic_max_hart_per_socket));
625         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-bits",
626             imsic_num_bits(socket_count));
627         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-shift",
628             IMSIC_MMIO_GROUP_MIN_SHIFT);
629     }
630     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "phandle", *msi_s_phandle);
631     g_free(imsic_name);
632 
633     g_free(imsic_regs);
634     g_free(imsic_cells);
635 }
636 
637 static void create_fdt_socket_aplic(RISCVVirtState *s,
638                                     const MemMapEntry *memmap, int socket,
639                                     uint32_t msi_m_phandle,
640                                     uint32_t msi_s_phandle,
641                                     uint32_t *phandle,
642                                     uint32_t *intc_phandles,
643                                     uint32_t *aplic_phandles)
644 {
645     int cpu;
646     char *aplic_name;
647     uint32_t *aplic_cells;
648     unsigned long aplic_addr;
649     MachineState *ms = MACHINE(s);
650     uint32_t aplic_m_phandle, aplic_s_phandle;
651 
652     aplic_m_phandle = (*phandle)++;
653     aplic_s_phandle = (*phandle)++;
654     aplic_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
655 
656     /* M-level APLIC node */
657     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
658         aplic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
659         aplic_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_M_EXT);
660     }
661     aplic_addr = memmap[VIRT_APLIC_M].base +
662                  (memmap[VIRT_APLIC_M].size * socket);
663     aplic_name = g_strdup_printf("/soc/aplic@%lx", aplic_addr);
664     qemu_fdt_add_subnode(ms->fdt, aplic_name);
665     qemu_fdt_setprop_string(ms->fdt, aplic_name, "compatible", "riscv,aplic");
666     qemu_fdt_setprop_cell(ms->fdt, aplic_name,
667         "#interrupt-cells", FDT_APLIC_INT_CELLS);
668     qemu_fdt_setprop(ms->fdt, aplic_name, "interrupt-controller", NULL, 0);
669     if (s->aia_type == VIRT_AIA_TYPE_APLIC) {
670         qemu_fdt_setprop(ms->fdt, aplic_name, "interrupts-extended",
671             aplic_cells, s->soc[socket].num_harts * sizeof(uint32_t) * 2);
672     } else {
673         qemu_fdt_setprop_cell(ms->fdt, aplic_name, "msi-parent",
674             msi_m_phandle);
675     }
676     qemu_fdt_setprop_cells(ms->fdt, aplic_name, "reg",
677         0x0, aplic_addr, 0x0, memmap[VIRT_APLIC_M].size);
678     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "riscv,num-sources",
679         VIRT_IRQCHIP_NUM_SOURCES);
680     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "riscv,children",
681         aplic_s_phandle);
682     qemu_fdt_setprop_cells(ms->fdt, aplic_name, "riscv,delegate",
683         aplic_s_phandle, 0x1, VIRT_IRQCHIP_NUM_SOURCES);
684     riscv_socket_fdt_write_id(ms, aplic_name, socket);
685     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "phandle", aplic_m_phandle);
686     g_free(aplic_name);
687 
688     /* S-level APLIC node */
689     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
690         aplic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
691         aplic_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_S_EXT);
692     }
693     aplic_addr = memmap[VIRT_APLIC_S].base +
694                  (memmap[VIRT_APLIC_S].size * socket);
695     aplic_name = g_strdup_printf("/soc/aplic@%lx", aplic_addr);
696     qemu_fdt_add_subnode(ms->fdt, aplic_name);
697     qemu_fdt_setprop_string(ms->fdt, aplic_name, "compatible", "riscv,aplic");
698     qemu_fdt_setprop_cell(ms->fdt, aplic_name,
699         "#interrupt-cells", FDT_APLIC_INT_CELLS);
700     qemu_fdt_setprop(ms->fdt, aplic_name, "interrupt-controller", NULL, 0);
701     if (s->aia_type == VIRT_AIA_TYPE_APLIC) {
702         qemu_fdt_setprop(ms->fdt, aplic_name, "interrupts-extended",
703             aplic_cells, s->soc[socket].num_harts * sizeof(uint32_t) * 2);
704     } else {
705         qemu_fdt_setprop_cell(ms->fdt, aplic_name, "msi-parent",
706             msi_s_phandle);
707     }
708     qemu_fdt_setprop_cells(ms->fdt, aplic_name, "reg",
709         0x0, aplic_addr, 0x0, memmap[VIRT_APLIC_S].size);
710     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "riscv,num-sources",
711         VIRT_IRQCHIP_NUM_SOURCES);
712     riscv_socket_fdt_write_id(ms, aplic_name, socket);
713     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "phandle", aplic_s_phandle);
714 
715     if (!socket) {
716         platform_bus_add_all_fdt_nodes(ms->fdt, aplic_name,
717                                        memmap[VIRT_PLATFORM_BUS].base,
718                                        memmap[VIRT_PLATFORM_BUS].size,
719                                        VIRT_PLATFORM_BUS_IRQ);
720     }
721 
722     g_free(aplic_name);
723 
724     g_free(aplic_cells);
725     aplic_phandles[socket] = aplic_s_phandle;
726 }
727 
728 static void create_fdt_pmu(RISCVVirtState *s)
729 {
730     char *pmu_name;
731     MachineState *ms = MACHINE(s);
732     RISCVCPU hart = s->soc[0].harts[0];
733 
734     pmu_name = g_strdup_printf("/soc/pmu");
735     qemu_fdt_add_subnode(ms->fdt, pmu_name);
736     qemu_fdt_setprop_string(ms->fdt, pmu_name, "compatible", "riscv,pmu");
737     riscv_pmu_generate_fdt_node(ms->fdt, hart.cfg.pmu_num, pmu_name);
738 
739     g_free(pmu_name);
740 }
741 
742 static void create_fdt_sockets(RISCVVirtState *s, const MemMapEntry *memmap,
743                                uint32_t *phandle,
744                                uint32_t *irq_mmio_phandle,
745                                uint32_t *irq_pcie_phandle,
746                                uint32_t *irq_virtio_phandle,
747                                uint32_t *msi_pcie_phandle)
748 {
749     char *clust_name;
750     int socket, phandle_pos;
751     MachineState *ms = MACHINE(s);
752     uint32_t msi_m_phandle = 0, msi_s_phandle = 0;
753     uint32_t *intc_phandles, xplic_phandles[MAX_NODES];
754     int socket_count = riscv_socket_count(ms);
755 
756     qemu_fdt_add_subnode(ms->fdt, "/cpus");
757     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "timebase-frequency",
758                           RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ);
759     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
760     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", 0x1);
761     qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
762 
763     intc_phandles = g_new0(uint32_t, ms->smp.cpus);
764 
765     phandle_pos = ms->smp.cpus;
766     for (socket = (socket_count - 1); socket >= 0; socket--) {
767         phandle_pos -= s->soc[socket].num_harts;
768 
769         clust_name = g_strdup_printf("/cpus/cpu-map/cluster%d", socket);
770         qemu_fdt_add_subnode(ms->fdt, clust_name);
771 
772         create_fdt_socket_cpus(s, socket, clust_name, phandle,
773                                &intc_phandles[phandle_pos]);
774 
775         create_fdt_socket_memory(s, memmap, socket);
776 
777         g_free(clust_name);
778 
779         if (!kvm_enabled()) {
780             if (s->have_aclint) {
781                 create_fdt_socket_aclint(s, memmap, socket,
782                     &intc_phandles[phandle_pos]);
783             } else {
784                 create_fdt_socket_clint(s, memmap, socket,
785                     &intc_phandles[phandle_pos]);
786             }
787         }
788     }
789 
790     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
791         create_fdt_imsic(s, memmap, phandle, intc_phandles,
792             &msi_m_phandle, &msi_s_phandle);
793         *msi_pcie_phandle = msi_s_phandle;
794     }
795 
796     phandle_pos = ms->smp.cpus;
797     for (socket = (socket_count - 1); socket >= 0; socket--) {
798         phandle_pos -= s->soc[socket].num_harts;
799 
800         if (s->aia_type == VIRT_AIA_TYPE_NONE) {
801             create_fdt_socket_plic(s, memmap, socket, phandle,
802                 &intc_phandles[phandle_pos], xplic_phandles);
803         } else {
804             create_fdt_socket_aplic(s, memmap, socket,
805                 msi_m_phandle, msi_s_phandle, phandle,
806                 &intc_phandles[phandle_pos], xplic_phandles);
807         }
808     }
809 
810     g_free(intc_phandles);
811 
812     for (socket = 0; socket < socket_count; socket++) {
813         if (socket == 0) {
814             *irq_mmio_phandle = xplic_phandles[socket];
815             *irq_virtio_phandle = xplic_phandles[socket];
816             *irq_pcie_phandle = xplic_phandles[socket];
817         }
818         if (socket == 1) {
819             *irq_virtio_phandle = xplic_phandles[socket];
820             *irq_pcie_phandle = xplic_phandles[socket];
821         }
822         if (socket == 2) {
823             *irq_pcie_phandle = xplic_phandles[socket];
824         }
825     }
826 
827     riscv_socket_fdt_write_distance_matrix(ms);
828 }
829 
830 static void create_fdt_virtio(RISCVVirtState *s, const MemMapEntry *memmap,
831                               uint32_t irq_virtio_phandle)
832 {
833     int i;
834     char *name;
835     MachineState *ms = MACHINE(s);
836 
837     for (i = 0; i < VIRTIO_COUNT; i++) {
838         name = g_strdup_printf("/soc/virtio_mmio@%lx",
839             (long)(memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size));
840         qemu_fdt_add_subnode(ms->fdt, name);
841         qemu_fdt_setprop_string(ms->fdt, name, "compatible", "virtio,mmio");
842         qemu_fdt_setprop_cells(ms->fdt, name, "reg",
843             0x0, memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size,
844             0x0, memmap[VIRT_VIRTIO].size);
845         qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent",
846             irq_virtio_phandle);
847         if (s->aia_type == VIRT_AIA_TYPE_NONE) {
848             qemu_fdt_setprop_cell(ms->fdt, name, "interrupts",
849                                   VIRTIO_IRQ + i);
850         } else {
851             qemu_fdt_setprop_cells(ms->fdt, name, "interrupts",
852                                    VIRTIO_IRQ + i, 0x4);
853         }
854         g_free(name);
855     }
856 }
857 
858 static void create_fdt_pcie(RISCVVirtState *s, const MemMapEntry *memmap,
859                             uint32_t irq_pcie_phandle,
860                             uint32_t msi_pcie_phandle)
861 {
862     char *name;
863     MachineState *ms = MACHINE(s);
864 
865     name = g_strdup_printf("/soc/pci@%lx",
866         (long) memmap[VIRT_PCIE_ECAM].base);
867     qemu_fdt_add_subnode(ms->fdt, name);
868     qemu_fdt_setprop_cell(ms->fdt, name, "#address-cells",
869         FDT_PCI_ADDR_CELLS);
870     qemu_fdt_setprop_cell(ms->fdt, name, "#interrupt-cells",
871         FDT_PCI_INT_CELLS);
872     qemu_fdt_setprop_cell(ms->fdt, name, "#size-cells", 0x2);
873     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
874         "pci-host-ecam-generic");
875     qemu_fdt_setprop_string(ms->fdt, name, "device_type", "pci");
876     qemu_fdt_setprop_cell(ms->fdt, name, "linux,pci-domain", 0);
877     qemu_fdt_setprop_cells(ms->fdt, name, "bus-range", 0,
878         memmap[VIRT_PCIE_ECAM].size / PCIE_MMCFG_SIZE_MIN - 1);
879     qemu_fdt_setprop(ms->fdt, name, "dma-coherent", NULL, 0);
880     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
881         qemu_fdt_setprop_cell(ms->fdt, name, "msi-parent", msi_pcie_phandle);
882     }
883     qemu_fdt_setprop_cells(ms->fdt, name, "reg", 0,
884         memmap[VIRT_PCIE_ECAM].base, 0, memmap[VIRT_PCIE_ECAM].size);
885     qemu_fdt_setprop_sized_cells(ms->fdt, name, "ranges",
886         1, FDT_PCI_RANGE_IOPORT, 2, 0,
887         2, memmap[VIRT_PCIE_PIO].base, 2, memmap[VIRT_PCIE_PIO].size,
888         1, FDT_PCI_RANGE_MMIO,
889         2, memmap[VIRT_PCIE_MMIO].base,
890         2, memmap[VIRT_PCIE_MMIO].base, 2, memmap[VIRT_PCIE_MMIO].size,
891         1, FDT_PCI_RANGE_MMIO_64BIT,
892         2, virt_high_pcie_memmap.base,
893         2, virt_high_pcie_memmap.base, 2, virt_high_pcie_memmap.size);
894 
895     create_pcie_irq_map(s, ms->fdt, name, irq_pcie_phandle);
896     g_free(name);
897 }
898 
899 static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap,
900                              uint32_t *phandle)
901 {
902     char *name;
903     uint32_t test_phandle;
904     MachineState *ms = MACHINE(s);
905 
906     test_phandle = (*phandle)++;
907     name = g_strdup_printf("/soc/test@%lx",
908         (long)memmap[VIRT_TEST].base);
909     qemu_fdt_add_subnode(ms->fdt, name);
910     {
911         static const char * const compat[3] = {
912             "sifive,test1", "sifive,test0", "syscon"
913         };
914         qemu_fdt_setprop_string_array(ms->fdt, name, "compatible",
915                                       (char **)&compat, ARRAY_SIZE(compat));
916     }
917     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
918         0x0, memmap[VIRT_TEST].base, 0x0, memmap[VIRT_TEST].size);
919     qemu_fdt_setprop_cell(ms->fdt, name, "phandle", test_phandle);
920     test_phandle = qemu_fdt_get_phandle(ms->fdt, name);
921     g_free(name);
922 
923     name = g_strdup_printf("/reboot");
924     qemu_fdt_add_subnode(ms->fdt, name);
925     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "syscon-reboot");
926     qemu_fdt_setprop_cell(ms->fdt, name, "regmap", test_phandle);
927     qemu_fdt_setprop_cell(ms->fdt, name, "offset", 0x0);
928     qemu_fdt_setprop_cell(ms->fdt, name, "value", FINISHER_RESET);
929     g_free(name);
930 
931     name = g_strdup_printf("/poweroff");
932     qemu_fdt_add_subnode(ms->fdt, name);
933     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "syscon-poweroff");
934     qemu_fdt_setprop_cell(ms->fdt, name, "regmap", test_phandle);
935     qemu_fdt_setprop_cell(ms->fdt, name, "offset", 0x0);
936     qemu_fdt_setprop_cell(ms->fdt, name, "value", FINISHER_PASS);
937     g_free(name);
938 }
939 
940 static void create_fdt_uart(RISCVVirtState *s, const MemMapEntry *memmap,
941                             uint32_t irq_mmio_phandle)
942 {
943     char *name;
944     MachineState *ms = MACHINE(s);
945 
946     name = g_strdup_printf("/soc/serial@%lx", (long)memmap[VIRT_UART0].base);
947     qemu_fdt_add_subnode(ms->fdt, name);
948     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "ns16550a");
949     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
950         0x0, memmap[VIRT_UART0].base,
951         0x0, memmap[VIRT_UART0].size);
952     qemu_fdt_setprop_cell(ms->fdt, name, "clock-frequency", 3686400);
953     qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent", irq_mmio_phandle);
954     if (s->aia_type == VIRT_AIA_TYPE_NONE) {
955         qemu_fdt_setprop_cell(ms->fdt, name, "interrupts", UART0_IRQ);
956     } else {
957         qemu_fdt_setprop_cells(ms->fdt, name, "interrupts", UART0_IRQ, 0x4);
958     }
959 
960     qemu_fdt_add_subnode(ms->fdt, "/chosen");
961     qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", name);
962     g_free(name);
963 }
964 
965 static void create_fdt_rtc(RISCVVirtState *s, const MemMapEntry *memmap,
966                            uint32_t irq_mmio_phandle)
967 {
968     char *name;
969     MachineState *ms = MACHINE(s);
970 
971     name = g_strdup_printf("/soc/rtc@%lx", (long)memmap[VIRT_RTC].base);
972     qemu_fdt_add_subnode(ms->fdt, name);
973     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
974         "google,goldfish-rtc");
975     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
976         0x0, memmap[VIRT_RTC].base, 0x0, memmap[VIRT_RTC].size);
977     qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent",
978         irq_mmio_phandle);
979     if (s->aia_type == VIRT_AIA_TYPE_NONE) {
980         qemu_fdt_setprop_cell(ms->fdt, name, "interrupts", RTC_IRQ);
981     } else {
982         qemu_fdt_setprop_cells(ms->fdt, name, "interrupts", RTC_IRQ, 0x4);
983     }
984     g_free(name);
985 }
986 
987 static void create_fdt_flash(RISCVVirtState *s, const MemMapEntry *memmap)
988 {
989     char *name;
990     MachineState *ms = MACHINE(s);
991     hwaddr flashsize = virt_memmap[VIRT_FLASH].size / 2;
992     hwaddr flashbase = virt_memmap[VIRT_FLASH].base;
993 
994     name = g_strdup_printf("/flash@%" PRIx64, flashbase);
995     qemu_fdt_add_subnode(ms->fdt, name);
996     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "cfi-flash");
997     qemu_fdt_setprop_sized_cells(ms->fdt, name, "reg",
998                                  2, flashbase, 2, flashsize,
999                                  2, flashbase + flashsize, 2, flashsize);
1000     qemu_fdt_setprop_cell(ms->fdt, name, "bank-width", 4);
1001     g_free(name);
1002 }
1003 
1004 static void create_fdt_fw_cfg(RISCVVirtState *s, const MemMapEntry *memmap)
1005 {
1006     char *nodename;
1007     MachineState *ms = MACHINE(s);
1008     hwaddr base = memmap[VIRT_FW_CFG].base;
1009     hwaddr size = memmap[VIRT_FW_CFG].size;
1010 
1011     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1012     qemu_fdt_add_subnode(ms->fdt, nodename);
1013     qemu_fdt_setprop_string(ms->fdt, nodename,
1014                             "compatible", "qemu,fw-cfg-mmio");
1015     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1016                                  2, base, 2, size);
1017     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1018     g_free(nodename);
1019 }
1020 
1021 static void create_fdt(RISCVVirtState *s, const MemMapEntry *memmap)
1022 {
1023     MachineState *ms = MACHINE(s);
1024     uint32_t phandle = 1, irq_mmio_phandle = 1, msi_pcie_phandle = 1;
1025     uint32_t irq_pcie_phandle = 1, irq_virtio_phandle = 1;
1026     uint8_t rng_seed[32];
1027 
1028     ms->fdt = create_device_tree(&s->fdt_size);
1029     if (!ms->fdt) {
1030         error_report("create_device_tree() failed");
1031         exit(1);
1032     }
1033 
1034     qemu_fdt_setprop_string(ms->fdt, "/", "model", "riscv-virtio,qemu");
1035     qemu_fdt_setprop_string(ms->fdt, "/", "compatible", "riscv-virtio");
1036     qemu_fdt_setprop_cell(ms->fdt, "/", "#size-cells", 0x2);
1037     qemu_fdt_setprop_cell(ms->fdt, "/", "#address-cells", 0x2);
1038 
1039     qemu_fdt_add_subnode(ms->fdt, "/soc");
1040     qemu_fdt_setprop(ms->fdt, "/soc", "ranges", NULL, 0);
1041     qemu_fdt_setprop_string(ms->fdt, "/soc", "compatible", "simple-bus");
1042     qemu_fdt_setprop_cell(ms->fdt, "/soc", "#size-cells", 0x2);
1043     qemu_fdt_setprop_cell(ms->fdt, "/soc", "#address-cells", 0x2);
1044 
1045     create_fdt_sockets(s, memmap, &phandle, &irq_mmio_phandle,
1046                        &irq_pcie_phandle, &irq_virtio_phandle,
1047                        &msi_pcie_phandle);
1048 
1049     create_fdt_virtio(s, memmap, irq_virtio_phandle);
1050 
1051     create_fdt_pcie(s, memmap, irq_pcie_phandle, msi_pcie_phandle);
1052 
1053     create_fdt_reset(s, memmap, &phandle);
1054 
1055     create_fdt_uart(s, memmap, irq_mmio_phandle);
1056 
1057     create_fdt_rtc(s, memmap, irq_mmio_phandle);
1058 
1059     create_fdt_flash(s, memmap);
1060     create_fdt_fw_cfg(s, memmap);
1061     create_fdt_pmu(s);
1062 
1063     /* Pass seed to RNG */
1064     qemu_guest_getrandom_nofail(rng_seed, sizeof(rng_seed));
1065     qemu_fdt_setprop(ms->fdt, "/chosen", "rng-seed",
1066                      rng_seed, sizeof(rng_seed));
1067 }
1068 
1069 static inline DeviceState *gpex_pcie_init(MemoryRegion *sys_mem,
1070                                           hwaddr ecam_base, hwaddr ecam_size,
1071                                           hwaddr mmio_base, hwaddr mmio_size,
1072                                           hwaddr high_mmio_base,
1073                                           hwaddr high_mmio_size,
1074                                           hwaddr pio_base,
1075                                           DeviceState *irqchip)
1076 {
1077     DeviceState *dev;
1078     MemoryRegion *ecam_alias, *ecam_reg;
1079     MemoryRegion *mmio_alias, *high_mmio_alias, *mmio_reg;
1080     qemu_irq irq;
1081     int i;
1082 
1083     dev = qdev_new(TYPE_GPEX_HOST);
1084 
1085     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1086 
1087     ecam_alias = g_new0(MemoryRegion, 1);
1088     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1089     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1090                              ecam_reg, 0, ecam_size);
1091     memory_region_add_subregion(get_system_memory(), ecam_base, ecam_alias);
1092 
1093     mmio_alias = g_new0(MemoryRegion, 1);
1094     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1095     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1096                              mmio_reg, mmio_base, mmio_size);
1097     memory_region_add_subregion(get_system_memory(), mmio_base, mmio_alias);
1098 
1099     /* Map high MMIO space */
1100     high_mmio_alias = g_new0(MemoryRegion, 1);
1101     memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1102                              mmio_reg, high_mmio_base, high_mmio_size);
1103     memory_region_add_subregion(get_system_memory(), high_mmio_base,
1104                                 high_mmio_alias);
1105 
1106     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, pio_base);
1107 
1108     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1109         irq = qdev_get_gpio_in(irqchip, PCIE_IRQ + i);
1110 
1111         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, irq);
1112         gpex_set_irq_num(GPEX_HOST(dev), i, PCIE_IRQ + i);
1113     }
1114 
1115     return dev;
1116 }
1117 
1118 static FWCfgState *create_fw_cfg(const MachineState *ms)
1119 {
1120     hwaddr base = virt_memmap[VIRT_FW_CFG].base;
1121     FWCfgState *fw_cfg;
1122 
1123     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16,
1124                                   &address_space_memory);
1125     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1126 
1127     return fw_cfg;
1128 }
1129 
1130 static DeviceState *virt_create_plic(const MemMapEntry *memmap, int socket,
1131                                      int base_hartid, int hart_count)
1132 {
1133     DeviceState *ret;
1134     char *plic_hart_config;
1135 
1136     /* Per-socket PLIC hart topology configuration string */
1137     plic_hart_config = riscv_plic_hart_config_string(hart_count);
1138 
1139     /* Per-socket PLIC */
1140     ret = sifive_plic_create(
1141             memmap[VIRT_PLIC].base + socket * memmap[VIRT_PLIC].size,
1142             plic_hart_config, hart_count, base_hartid,
1143             VIRT_IRQCHIP_NUM_SOURCES,
1144             ((1U << VIRT_IRQCHIP_NUM_PRIO_BITS) - 1),
1145             VIRT_PLIC_PRIORITY_BASE,
1146             VIRT_PLIC_PENDING_BASE,
1147             VIRT_PLIC_ENABLE_BASE,
1148             VIRT_PLIC_ENABLE_STRIDE,
1149             VIRT_PLIC_CONTEXT_BASE,
1150             VIRT_PLIC_CONTEXT_STRIDE,
1151             memmap[VIRT_PLIC].size);
1152 
1153     g_free(plic_hart_config);
1154 
1155     return ret;
1156 }
1157 
1158 static DeviceState *virt_create_aia(RISCVVirtAIAType aia_type, int aia_guests,
1159                                     const MemMapEntry *memmap, int socket,
1160                                     int base_hartid, int hart_count)
1161 {
1162     int i;
1163     hwaddr addr;
1164     uint32_t guest_bits;
1165     DeviceState *aplic_m;
1166     bool msimode = (aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) ? true : false;
1167 
1168     if (msimode) {
1169         /* Per-socket M-level IMSICs */
1170         addr = memmap[VIRT_IMSIC_M].base + socket * VIRT_IMSIC_GROUP_MAX_SIZE;
1171         for (i = 0; i < hart_count; i++) {
1172             riscv_imsic_create(addr + i * IMSIC_HART_SIZE(0),
1173                                base_hartid + i, true, 1,
1174                                VIRT_IRQCHIP_NUM_MSIS);
1175         }
1176 
1177         /* Per-socket S-level IMSICs */
1178         guest_bits = imsic_num_bits(aia_guests + 1);
1179         addr = memmap[VIRT_IMSIC_S].base + socket * VIRT_IMSIC_GROUP_MAX_SIZE;
1180         for (i = 0; i < hart_count; i++) {
1181             riscv_imsic_create(addr + i * IMSIC_HART_SIZE(guest_bits),
1182                                base_hartid + i, false, 1 + aia_guests,
1183                                VIRT_IRQCHIP_NUM_MSIS);
1184         }
1185     }
1186 
1187     /* Per-socket M-level APLIC */
1188     aplic_m = riscv_aplic_create(
1189         memmap[VIRT_APLIC_M].base + socket * memmap[VIRT_APLIC_M].size,
1190         memmap[VIRT_APLIC_M].size,
1191         (msimode) ? 0 : base_hartid,
1192         (msimode) ? 0 : hart_count,
1193         VIRT_IRQCHIP_NUM_SOURCES,
1194         VIRT_IRQCHIP_NUM_PRIO_BITS,
1195         msimode, true, NULL);
1196 
1197     if (aplic_m) {
1198         /* Per-socket S-level APLIC */
1199         riscv_aplic_create(
1200             memmap[VIRT_APLIC_S].base + socket * memmap[VIRT_APLIC_S].size,
1201             memmap[VIRT_APLIC_S].size,
1202             (msimode) ? 0 : base_hartid,
1203             (msimode) ? 0 : hart_count,
1204             VIRT_IRQCHIP_NUM_SOURCES,
1205             VIRT_IRQCHIP_NUM_PRIO_BITS,
1206             msimode, false, aplic_m);
1207     }
1208 
1209     return aplic_m;
1210 }
1211 
1212 static void create_platform_bus(RISCVVirtState *s, DeviceState *irqchip)
1213 {
1214     DeviceState *dev;
1215     SysBusDevice *sysbus;
1216     const MemMapEntry *memmap = virt_memmap;
1217     int i;
1218     MemoryRegion *sysmem = get_system_memory();
1219 
1220     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1221     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1222     qdev_prop_set_uint32(dev, "num_irqs", VIRT_PLATFORM_BUS_NUM_IRQS);
1223     qdev_prop_set_uint32(dev, "mmio_size", memmap[VIRT_PLATFORM_BUS].size);
1224     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1225     s->platform_bus_dev = dev;
1226 
1227     sysbus = SYS_BUS_DEVICE(dev);
1228     for (i = 0; i < VIRT_PLATFORM_BUS_NUM_IRQS; i++) {
1229         int irq = VIRT_PLATFORM_BUS_IRQ + i;
1230         sysbus_connect_irq(sysbus, i, qdev_get_gpio_in(irqchip, irq));
1231     }
1232 
1233     memory_region_add_subregion(sysmem,
1234                                 memmap[VIRT_PLATFORM_BUS].base,
1235                                 sysbus_mmio_get_region(sysbus, 0));
1236 }
1237 
1238 static void virt_machine_done(Notifier *notifier, void *data)
1239 {
1240     RISCVVirtState *s = container_of(notifier, RISCVVirtState,
1241                                      machine_done);
1242     const MemMapEntry *memmap = virt_memmap;
1243     MachineState *machine = MACHINE(s);
1244     target_ulong start_addr = memmap[VIRT_DRAM].base;
1245     target_ulong firmware_end_addr, kernel_start_addr;
1246     const char *firmware_name = riscv_default_firmware_name(&s->soc[0]);
1247     uint32_t fdt_load_addr;
1248     uint64_t kernel_entry;
1249 
1250     /*
1251      * Only direct boot kernel is currently supported for KVM VM,
1252      * so the "-bios" parameter is not supported when KVM is enabled.
1253      */
1254     if (kvm_enabled()) {
1255         if (machine->firmware) {
1256             if (strcmp(machine->firmware, "none")) {
1257                 error_report("Machine mode firmware is not supported in "
1258                              "combination with KVM.");
1259                 exit(1);
1260             }
1261         } else {
1262             machine->firmware = g_strdup("none");
1263         }
1264     }
1265 
1266     firmware_end_addr = riscv_find_and_load_firmware(machine, firmware_name,
1267                                                      start_addr, NULL);
1268 
1269     if (drive_get(IF_PFLASH, 0, 1)) {
1270         /*
1271          * S-mode FW like EDK2 will be kept in second plash (unit 1).
1272          * When both kernel, initrd and pflash options are provided in the
1273          * command line, the kernel and initrd will be copied to the fw_cfg
1274          * table and opensbi will jump to the flash address which is the
1275          * entry point of S-mode FW. It is the job of the S-mode FW to load
1276          * the kernel and initrd using fw_cfg table.
1277          *
1278          * If only pflash is given but not -kernel, then it is the job of
1279          * of the S-mode firmware to locate and load the kernel.
1280          * In either case, the next_addr for opensbi will be the flash address.
1281          */
1282         riscv_setup_firmware_boot(machine);
1283         kernel_entry = virt_memmap[VIRT_FLASH].base +
1284                        virt_memmap[VIRT_FLASH].size / 2;
1285     } else if (machine->kernel_filename) {
1286         kernel_start_addr = riscv_calc_kernel_start_addr(&s->soc[0],
1287                                                          firmware_end_addr);
1288 
1289         kernel_entry = riscv_load_kernel(machine, &s->soc[0],
1290                                          kernel_start_addr, true, NULL);
1291     } else {
1292        /*
1293         * If dynamic firmware is used, it doesn't know where is the next mode
1294         * if kernel argument is not set.
1295         */
1296         kernel_entry = 0;
1297     }
1298 
1299     if (drive_get(IF_PFLASH, 0, 0)) {
1300         /*
1301          * Pflash was supplied, let's overwrite the address we jump to after
1302          * reset to the base of the flash.
1303          */
1304         start_addr = virt_memmap[VIRT_FLASH].base;
1305     }
1306 
1307     fdt_load_addr = riscv_compute_fdt_addr(memmap[VIRT_DRAM].base,
1308                                            memmap[VIRT_DRAM].size,
1309                                            machine);
1310     riscv_load_fdt(fdt_load_addr, machine->fdt);
1311 
1312     /* load the reset vector */
1313     riscv_setup_rom_reset_vec(machine, &s->soc[0], start_addr,
1314                               virt_memmap[VIRT_MROM].base,
1315                               virt_memmap[VIRT_MROM].size, kernel_entry,
1316                               fdt_load_addr);
1317 
1318     /*
1319      * Only direct boot kernel is currently supported for KVM VM,
1320      * So here setup kernel start address and fdt address.
1321      * TODO:Support firmware loading and integrate to TCG start
1322      */
1323     if (kvm_enabled()) {
1324         riscv_setup_direct_kernel(kernel_entry, fdt_load_addr);
1325     }
1326 
1327     if (virt_is_acpi_enabled(s)) {
1328         virt_acpi_setup(s);
1329     }
1330 }
1331 
1332 static void virt_machine_init(MachineState *machine)
1333 {
1334     const MemMapEntry *memmap = virt_memmap;
1335     RISCVVirtState *s = RISCV_VIRT_MACHINE(machine);
1336     MemoryRegion *system_memory = get_system_memory();
1337     MemoryRegion *mask_rom = g_new(MemoryRegion, 1);
1338     char *soc_name;
1339     DeviceState *mmio_irqchip, *virtio_irqchip, *pcie_irqchip;
1340     int i, base_hartid, hart_count;
1341     int socket_count = riscv_socket_count(machine);
1342 
1343     /* Check socket count limit */
1344     if (VIRT_SOCKETS_MAX < socket_count) {
1345         error_report("number of sockets/nodes should be less than %d",
1346             VIRT_SOCKETS_MAX);
1347         exit(1);
1348     }
1349 
1350     /* Initialize sockets */
1351     mmio_irqchip = virtio_irqchip = pcie_irqchip = NULL;
1352     for (i = 0; i < socket_count; i++) {
1353         if (!riscv_socket_check_hartids(machine, i)) {
1354             error_report("discontinuous hartids in socket%d", i);
1355             exit(1);
1356         }
1357 
1358         base_hartid = riscv_socket_first_hartid(machine, i);
1359         if (base_hartid < 0) {
1360             error_report("can't find hartid base for socket%d", i);
1361             exit(1);
1362         }
1363 
1364         hart_count = riscv_socket_hart_count(machine, i);
1365         if (hart_count < 0) {
1366             error_report("can't find hart count for socket%d", i);
1367             exit(1);
1368         }
1369 
1370         soc_name = g_strdup_printf("soc%d", i);
1371         object_initialize_child(OBJECT(machine), soc_name, &s->soc[i],
1372                                 TYPE_RISCV_HART_ARRAY);
1373         g_free(soc_name);
1374         object_property_set_str(OBJECT(&s->soc[i]), "cpu-type",
1375                                 machine->cpu_type, &error_abort);
1376         object_property_set_int(OBJECT(&s->soc[i]), "hartid-base",
1377                                 base_hartid, &error_abort);
1378         object_property_set_int(OBJECT(&s->soc[i]), "num-harts",
1379                                 hart_count, &error_abort);
1380         sysbus_realize(SYS_BUS_DEVICE(&s->soc[i]), &error_fatal);
1381 
1382         if (!kvm_enabled()) {
1383             if (s->have_aclint) {
1384                 if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
1385                     /* Per-socket ACLINT MTIMER */
1386                     riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1387                             i * RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1388                         RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1389                         base_hartid, hart_count,
1390                         RISCV_ACLINT_DEFAULT_MTIMECMP,
1391                         RISCV_ACLINT_DEFAULT_MTIME,
1392                         RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1393                 } else {
1394                     /* Per-socket ACLINT MSWI, MTIMER, and SSWI */
1395                     riscv_aclint_swi_create(memmap[VIRT_CLINT].base +
1396                             i * memmap[VIRT_CLINT].size,
1397                         base_hartid, hart_count, false);
1398                     riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1399                             i * memmap[VIRT_CLINT].size +
1400                             RISCV_ACLINT_SWI_SIZE,
1401                         RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1402                         base_hartid, hart_count,
1403                         RISCV_ACLINT_DEFAULT_MTIMECMP,
1404                         RISCV_ACLINT_DEFAULT_MTIME,
1405                         RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1406                     riscv_aclint_swi_create(memmap[VIRT_ACLINT_SSWI].base +
1407                             i * memmap[VIRT_ACLINT_SSWI].size,
1408                         base_hartid, hart_count, true);
1409                 }
1410             } else {
1411                 /* Per-socket SiFive CLINT */
1412                 riscv_aclint_swi_create(
1413                     memmap[VIRT_CLINT].base + i * memmap[VIRT_CLINT].size,
1414                     base_hartid, hart_count, false);
1415                 riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1416                         i * memmap[VIRT_CLINT].size + RISCV_ACLINT_SWI_SIZE,
1417                     RISCV_ACLINT_DEFAULT_MTIMER_SIZE, base_hartid, hart_count,
1418                     RISCV_ACLINT_DEFAULT_MTIMECMP, RISCV_ACLINT_DEFAULT_MTIME,
1419                     RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1420             }
1421         }
1422 
1423         /* Per-socket interrupt controller */
1424         if (s->aia_type == VIRT_AIA_TYPE_NONE) {
1425             s->irqchip[i] = virt_create_plic(memmap, i,
1426                                              base_hartid, hart_count);
1427         } else {
1428             s->irqchip[i] = virt_create_aia(s->aia_type, s->aia_guests,
1429                                             memmap, i, base_hartid,
1430                                             hart_count);
1431         }
1432 
1433         /* Try to use different IRQCHIP instance based device type */
1434         if (i == 0) {
1435             mmio_irqchip = s->irqchip[i];
1436             virtio_irqchip = s->irqchip[i];
1437             pcie_irqchip = s->irqchip[i];
1438         }
1439         if (i == 1) {
1440             virtio_irqchip = s->irqchip[i];
1441             pcie_irqchip = s->irqchip[i];
1442         }
1443         if (i == 2) {
1444             pcie_irqchip = s->irqchip[i];
1445         }
1446     }
1447 
1448     if (riscv_is_32bit(&s->soc[0])) {
1449 #if HOST_LONG_BITS == 64
1450         /* limit RAM size in a 32-bit system */
1451         if (machine->ram_size > 10 * GiB) {
1452             machine->ram_size = 10 * GiB;
1453             error_report("Limiting RAM size to 10 GiB");
1454         }
1455 #endif
1456         virt_high_pcie_memmap.base = VIRT32_HIGH_PCIE_MMIO_BASE;
1457         virt_high_pcie_memmap.size = VIRT32_HIGH_PCIE_MMIO_SIZE;
1458     } else {
1459         virt_high_pcie_memmap.size = VIRT64_HIGH_PCIE_MMIO_SIZE;
1460         virt_high_pcie_memmap.base = memmap[VIRT_DRAM].base + machine->ram_size;
1461         virt_high_pcie_memmap.base =
1462             ROUND_UP(virt_high_pcie_memmap.base, virt_high_pcie_memmap.size);
1463     }
1464 
1465     s->memmap = virt_memmap;
1466 
1467     /* register system main memory (actual RAM) */
1468     memory_region_add_subregion(system_memory, memmap[VIRT_DRAM].base,
1469         machine->ram);
1470 
1471     /* boot rom */
1472     memory_region_init_rom(mask_rom, NULL, "riscv_virt_board.mrom",
1473                            memmap[VIRT_MROM].size, &error_fatal);
1474     memory_region_add_subregion(system_memory, memmap[VIRT_MROM].base,
1475                                 mask_rom);
1476 
1477     /*
1478      * Init fw_cfg. Must be done before riscv_load_fdt, otherwise the
1479      * device tree cannot be altered and we get FDT_ERR_NOSPACE.
1480      */
1481     s->fw_cfg = create_fw_cfg(machine);
1482     rom_set_fw(s->fw_cfg);
1483 
1484     /* SiFive Test MMIO device */
1485     sifive_test_create(memmap[VIRT_TEST].base);
1486 
1487     /* VirtIO MMIO devices */
1488     for (i = 0; i < VIRTIO_COUNT; i++) {
1489         sysbus_create_simple("virtio-mmio",
1490             memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size,
1491             qdev_get_gpio_in(DEVICE(virtio_irqchip), VIRTIO_IRQ + i));
1492     }
1493 
1494     gpex_pcie_init(system_memory,
1495                    memmap[VIRT_PCIE_ECAM].base,
1496                    memmap[VIRT_PCIE_ECAM].size,
1497                    memmap[VIRT_PCIE_MMIO].base,
1498                    memmap[VIRT_PCIE_MMIO].size,
1499                    virt_high_pcie_memmap.base,
1500                    virt_high_pcie_memmap.size,
1501                    memmap[VIRT_PCIE_PIO].base,
1502                    DEVICE(pcie_irqchip));
1503 
1504     create_platform_bus(s, DEVICE(mmio_irqchip));
1505 
1506     serial_mm_init(system_memory, memmap[VIRT_UART0].base,
1507         0, qdev_get_gpio_in(DEVICE(mmio_irqchip), UART0_IRQ), 399193,
1508         serial_hd(0), DEVICE_LITTLE_ENDIAN);
1509 
1510     sysbus_create_simple("goldfish_rtc", memmap[VIRT_RTC].base,
1511         qdev_get_gpio_in(DEVICE(mmio_irqchip), RTC_IRQ));
1512 
1513     virt_flash_create(s);
1514 
1515     for (i = 0; i < ARRAY_SIZE(s->flash); i++) {
1516         /* Map legacy -drive if=pflash to machine properties */
1517         pflash_cfi01_legacy_drive(s->flash[i],
1518                                   drive_get(IF_PFLASH, 0, i));
1519     }
1520     virt_flash_map(s, system_memory);
1521 
1522     /* load/create device tree */
1523     if (machine->dtb) {
1524         machine->fdt = load_device_tree(machine->dtb, &s->fdt_size);
1525         if (!machine->fdt) {
1526             error_report("load_device_tree() failed");
1527             exit(1);
1528         }
1529     } else {
1530         create_fdt(s, memmap);
1531     }
1532 
1533     s->machine_done.notify = virt_machine_done;
1534     qemu_add_machine_init_done_notifier(&s->machine_done);
1535 }
1536 
1537 static void virt_machine_instance_init(Object *obj)
1538 {
1539     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1540 
1541     s->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
1542     s->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
1543     s->acpi = ON_OFF_AUTO_AUTO;
1544 }
1545 
1546 static char *virt_get_aia_guests(Object *obj, Error **errp)
1547 {
1548     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1549     char val[32];
1550 
1551     sprintf(val, "%d", s->aia_guests);
1552     return g_strdup(val);
1553 }
1554 
1555 static void virt_set_aia_guests(Object *obj, const char *val, Error **errp)
1556 {
1557     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1558 
1559     s->aia_guests = atoi(val);
1560     if (s->aia_guests < 0 || s->aia_guests > VIRT_IRQCHIP_MAX_GUESTS) {
1561         error_setg(errp, "Invalid number of AIA IMSIC guests");
1562         error_append_hint(errp, "Valid values be between 0 and %d.\n",
1563                           VIRT_IRQCHIP_MAX_GUESTS);
1564     }
1565 }
1566 
1567 static char *virt_get_aia(Object *obj, Error **errp)
1568 {
1569     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1570     const char *val;
1571 
1572     switch (s->aia_type) {
1573     case VIRT_AIA_TYPE_APLIC:
1574         val = "aplic";
1575         break;
1576     case VIRT_AIA_TYPE_APLIC_IMSIC:
1577         val = "aplic-imsic";
1578         break;
1579     default:
1580         val = "none";
1581         break;
1582     };
1583 
1584     return g_strdup(val);
1585 }
1586 
1587 static void virt_set_aia(Object *obj, const char *val, Error **errp)
1588 {
1589     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1590 
1591     if (!strcmp(val, "none")) {
1592         s->aia_type = VIRT_AIA_TYPE_NONE;
1593     } else if (!strcmp(val, "aplic")) {
1594         s->aia_type = VIRT_AIA_TYPE_APLIC;
1595     } else if (!strcmp(val, "aplic-imsic")) {
1596         s->aia_type = VIRT_AIA_TYPE_APLIC_IMSIC;
1597     } else {
1598         error_setg(errp, "Invalid AIA interrupt controller type");
1599         error_append_hint(errp, "Valid values are none, aplic, and "
1600                           "aplic-imsic.\n");
1601     }
1602 }
1603 
1604 static bool virt_get_aclint(Object *obj, Error **errp)
1605 {
1606     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1607 
1608     return s->have_aclint;
1609 }
1610 
1611 static void virt_set_aclint(Object *obj, bool value, Error **errp)
1612 {
1613     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1614 
1615     s->have_aclint = value;
1616 }
1617 
1618 bool virt_is_acpi_enabled(RISCVVirtState *s)
1619 {
1620     return s->acpi != ON_OFF_AUTO_OFF;
1621 }
1622 
1623 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
1624                           void *opaque, Error **errp)
1625 {
1626     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1627     OnOffAuto acpi = s->acpi;
1628 
1629     visit_type_OnOffAuto(v, name, &acpi, errp);
1630 }
1631 
1632 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
1633                           void *opaque, Error **errp)
1634 {
1635     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1636 
1637     visit_type_OnOffAuto(v, name, &s->acpi, errp);
1638 }
1639 
1640 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
1641                                                         DeviceState *dev)
1642 {
1643     MachineClass *mc = MACHINE_GET_CLASS(machine);
1644 
1645     if (device_is_dynamic_sysbus(mc, dev)) {
1646         return HOTPLUG_HANDLER(machine);
1647     }
1648     return NULL;
1649 }
1650 
1651 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1652                                         DeviceState *dev, Error **errp)
1653 {
1654     RISCVVirtState *s = RISCV_VIRT_MACHINE(hotplug_dev);
1655 
1656     if (s->platform_bus_dev) {
1657         MachineClass *mc = MACHINE_GET_CLASS(s);
1658 
1659         if (device_is_dynamic_sysbus(mc, dev)) {
1660             platform_bus_link_device(PLATFORM_BUS_DEVICE(s->platform_bus_dev),
1661                                      SYS_BUS_DEVICE(dev));
1662         }
1663     }
1664 }
1665 
1666 static void virt_machine_class_init(ObjectClass *oc, void *data)
1667 {
1668     char str[128];
1669     MachineClass *mc = MACHINE_CLASS(oc);
1670     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1671 
1672     mc->desc = "RISC-V VirtIO board";
1673     mc->init = virt_machine_init;
1674     mc->max_cpus = VIRT_CPUS_MAX;
1675     mc->default_cpu_type = TYPE_RISCV_CPU_BASE;
1676     mc->pci_allow_0_address = true;
1677     mc->possible_cpu_arch_ids = riscv_numa_possible_cpu_arch_ids;
1678     mc->cpu_index_to_instance_props = riscv_numa_cpu_index_to_props;
1679     mc->get_default_cpu_node_id = riscv_numa_get_default_cpu_node_id;
1680     mc->numa_mem_supported = true;
1681     mc->default_ram_id = "riscv_virt_board.ram";
1682     assert(!mc->get_hotplug_handler);
1683     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
1684 
1685     hc->plug = virt_machine_device_plug_cb;
1686 
1687     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
1688 #ifdef CONFIG_TPM
1689     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
1690 #endif
1691 
1692     object_class_property_add_bool(oc, "aclint", virt_get_aclint,
1693                                    virt_set_aclint);
1694     object_class_property_set_description(oc, "aclint",
1695                                           "Set on/off to enable/disable "
1696                                           "emulating ACLINT devices");
1697 
1698     object_class_property_add_str(oc, "aia", virt_get_aia,
1699                                   virt_set_aia);
1700     object_class_property_set_description(oc, "aia",
1701                                           "Set type of AIA interrupt "
1702                                           "conttoller. Valid values are "
1703                                           "none, aplic, and aplic-imsic.");
1704 
1705     object_class_property_add_str(oc, "aia-guests",
1706                                   virt_get_aia_guests,
1707                                   virt_set_aia_guests);
1708     sprintf(str, "Set number of guest MMIO pages for AIA IMSIC. Valid value "
1709                  "should be between 0 and %d.", VIRT_IRQCHIP_MAX_GUESTS);
1710     object_class_property_set_description(oc, "aia-guests", str);
1711     object_class_property_add(oc, "acpi", "OnOffAuto",
1712                               virt_get_acpi, virt_set_acpi,
1713                               NULL, NULL);
1714     object_class_property_set_description(oc, "acpi",
1715                                           "Enable ACPI");
1716 }
1717 
1718 static const TypeInfo virt_machine_typeinfo = {
1719     .name       = MACHINE_TYPE_NAME("virt"),
1720     .parent     = TYPE_MACHINE,
1721     .class_init = virt_machine_class_init,
1722     .instance_init = virt_machine_instance_init,
1723     .instance_size = sizeof(RISCVVirtState),
1724     .interfaces = (InterfaceInfo[]) {
1725          { TYPE_HOTPLUG_HANDLER },
1726          { }
1727     },
1728 };
1729 
1730 static void virt_machine_init_register_types(void)
1731 {
1732     type_register_static(&virt_machine_typeinfo);
1733 }
1734 
1735 type_init(virt_machine_init_register_types)
1736