xref: /openbmc/qemu/hw/riscv/virt.c (revision f81198ce)
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/firmware/smbios.h"
40 #include "hw/intc/riscv_aclint.h"
41 #include "hw/intc/riscv_aplic.h"
42 #include "hw/intc/sifive_plic.h"
43 #include "hw/misc/sifive_test.h"
44 #include "hw/platform-bus.h"
45 #include "chardev/char.h"
46 #include "sysemu/device_tree.h"
47 #include "sysemu/sysemu.h"
48 #include "sysemu/tcg.h"
49 #include "sysemu/kvm.h"
50 #include "sysemu/tpm.h"
51 #include "sysemu/qtest.h"
52 #include "hw/pci/pci.h"
53 #include "hw/pci-host/gpex.h"
54 #include "hw/display/ramfb.h"
55 #include "hw/acpi/aml-build.h"
56 #include "qapi/qapi-visit-common.h"
57 #include "hw/virtio/virtio-iommu.h"
58 
59 /* KVM AIA only supports APLIC MSI. APLIC Wired is always emulated by QEMU. */
60 static bool virt_use_kvm_aia(RISCVVirtState *s)
61 {
62     return kvm_irqchip_in_kernel() && s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC;
63 }
64 
65 static bool virt_aclint_allowed(void)
66 {
67     return tcg_enabled() || qtest_enabled();
68 }
69 
70 static const MemMapEntry virt_memmap[] = {
71     [VIRT_DEBUG] =        {        0x0,         0x100 },
72     [VIRT_MROM] =         {     0x1000,        0xf000 },
73     [VIRT_TEST] =         {   0x100000,        0x1000 },
74     [VIRT_RTC] =          {   0x101000,        0x1000 },
75     [VIRT_CLINT] =        {  0x2000000,       0x10000 },
76     [VIRT_ACLINT_SSWI] =  {  0x2F00000,        0x4000 },
77     [VIRT_PCIE_PIO] =     {  0x3000000,       0x10000 },
78     [VIRT_PLATFORM_BUS] = {  0x4000000,     0x2000000 },
79     [VIRT_PLIC] =         {  0xc000000, VIRT_PLIC_SIZE(VIRT_CPUS_MAX * 2) },
80     [VIRT_APLIC_M] =      {  0xc000000, APLIC_SIZE(VIRT_CPUS_MAX) },
81     [VIRT_APLIC_S] =      {  0xd000000, APLIC_SIZE(VIRT_CPUS_MAX) },
82     [VIRT_UART0] =        { 0x10000000,         0x100 },
83     [VIRT_VIRTIO] =       { 0x10001000,        0x1000 },
84     [VIRT_FW_CFG] =       { 0x10100000,          0x18 },
85     [VIRT_FLASH] =        { 0x20000000,     0x4000000 },
86     [VIRT_IMSIC_M] =      { 0x24000000, VIRT_IMSIC_MAX_SIZE },
87     [VIRT_IMSIC_S] =      { 0x28000000, VIRT_IMSIC_MAX_SIZE },
88     [VIRT_PCIE_ECAM] =    { 0x30000000,    0x10000000 },
89     [VIRT_PCIE_MMIO] =    { 0x40000000,    0x40000000 },
90     [VIRT_DRAM] =         { 0x80000000,           0x0 },
91 };
92 
93 /* PCIe high mmio is fixed for RV32 */
94 #define VIRT32_HIGH_PCIE_MMIO_BASE  0x300000000ULL
95 #define VIRT32_HIGH_PCIE_MMIO_SIZE  (4 * GiB)
96 
97 /* PCIe high mmio for RV64, size is fixed but base depends on top of RAM */
98 #define VIRT64_HIGH_PCIE_MMIO_SIZE  (16 * GiB)
99 
100 static MemMapEntry virt_high_pcie_memmap;
101 
102 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
103 
104 static PFlashCFI01 *virt_flash_create1(RISCVVirtState *s,
105                                        const char *name,
106                                        const char *alias_prop_name)
107 {
108     /*
109      * Create a single flash device.  We use the same parameters as
110      * the flash devices on the ARM virt board.
111      */
112     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
113 
114     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
115     qdev_prop_set_uint8(dev, "width", 4);
116     qdev_prop_set_uint8(dev, "device-width", 2);
117     qdev_prop_set_bit(dev, "big-endian", false);
118     qdev_prop_set_uint16(dev, "id0", 0x89);
119     qdev_prop_set_uint16(dev, "id1", 0x18);
120     qdev_prop_set_uint16(dev, "id2", 0x00);
121     qdev_prop_set_uint16(dev, "id3", 0x00);
122     qdev_prop_set_string(dev, "name", name);
123 
124     object_property_add_child(OBJECT(s), name, OBJECT(dev));
125     object_property_add_alias(OBJECT(s), alias_prop_name,
126                               OBJECT(dev), "drive");
127 
128     return PFLASH_CFI01(dev);
129 }
130 
131 static void virt_flash_create(RISCVVirtState *s)
132 {
133     s->flash[0] = virt_flash_create1(s, "virt.flash0", "pflash0");
134     s->flash[1] = virt_flash_create1(s, "virt.flash1", "pflash1");
135 }
136 
137 static void virt_flash_map1(PFlashCFI01 *flash,
138                             hwaddr base, hwaddr size,
139                             MemoryRegion *sysmem)
140 {
141     DeviceState *dev = DEVICE(flash);
142 
143     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
144     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
145     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
146     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
147 
148     memory_region_add_subregion(sysmem, base,
149                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
150                                                        0));
151 }
152 
153 static void virt_flash_map(RISCVVirtState *s,
154                            MemoryRegion *sysmem)
155 {
156     hwaddr flashsize = virt_memmap[VIRT_FLASH].size / 2;
157     hwaddr flashbase = virt_memmap[VIRT_FLASH].base;
158 
159     virt_flash_map1(s->flash[0], flashbase, flashsize,
160                     sysmem);
161     virt_flash_map1(s->flash[1], flashbase + flashsize, flashsize,
162                     sysmem);
163 }
164 
165 static void create_pcie_irq_map(RISCVVirtState *s, void *fdt, char *nodename,
166                                 uint32_t irqchip_phandle)
167 {
168     int pin, dev;
169     uint32_t irq_map_stride = 0;
170     uint32_t full_irq_map[GPEX_NUM_IRQS * GPEX_NUM_IRQS *
171                           FDT_MAX_INT_MAP_WIDTH] = {};
172     uint32_t *irq_map = full_irq_map;
173 
174     /* This code creates a standard swizzle of interrupts such that
175      * each device's first interrupt is based on it's PCI_SLOT number.
176      * (See pci_swizzle_map_irq_fn())
177      *
178      * We only need one entry per interrupt in the table (not one per
179      * possible slot) seeing the interrupt-map-mask will allow the table
180      * to wrap to any number of devices.
181      */
182     for (dev = 0; dev < GPEX_NUM_IRQS; dev++) {
183         int devfn = dev * 0x8;
184 
185         for (pin = 0; pin < GPEX_NUM_IRQS; pin++) {
186             int irq_nr = PCIE_IRQ + ((pin + PCI_SLOT(devfn)) % GPEX_NUM_IRQS);
187             int i = 0;
188 
189             /* Fill PCI address cells */
190             irq_map[i] = cpu_to_be32(devfn << 8);
191             i += FDT_PCI_ADDR_CELLS;
192 
193             /* Fill PCI Interrupt cells */
194             irq_map[i] = cpu_to_be32(pin + 1);
195             i += FDT_PCI_INT_CELLS;
196 
197             /* Fill interrupt controller phandle and cells */
198             irq_map[i++] = cpu_to_be32(irqchip_phandle);
199             irq_map[i++] = cpu_to_be32(irq_nr);
200             if (s->aia_type != VIRT_AIA_TYPE_NONE) {
201                 irq_map[i++] = cpu_to_be32(0x4);
202             }
203 
204             if (!irq_map_stride) {
205                 irq_map_stride = i;
206             }
207             irq_map += irq_map_stride;
208         }
209     }
210 
211     qemu_fdt_setprop(fdt, nodename, "interrupt-map", full_irq_map,
212                      GPEX_NUM_IRQS * GPEX_NUM_IRQS *
213                      irq_map_stride * sizeof(uint32_t));
214 
215     qemu_fdt_setprop_cells(fdt, nodename, "interrupt-map-mask",
216                            0x1800, 0, 0, 0x7);
217 }
218 
219 static void create_fdt_socket_cpus(RISCVVirtState *s, int socket,
220                                    char *clust_name, uint32_t *phandle,
221                                    uint32_t *intc_phandles)
222 {
223     int cpu;
224     uint32_t cpu_phandle;
225     MachineState *ms = MACHINE(s);
226     bool is_32_bit = riscv_is_32bit(&s->soc[0]);
227     uint8_t satp_mode_max;
228 
229     for (cpu = s->soc[socket].num_harts - 1; cpu >= 0; cpu--) {
230         RISCVCPU *cpu_ptr = &s->soc[socket].harts[cpu];
231         g_autofree char *cpu_name = NULL;
232         g_autofree char *core_name = NULL;
233         g_autofree char *intc_name = NULL;
234         g_autofree char *sv_name = NULL;
235 
236         cpu_phandle = (*phandle)++;
237 
238         cpu_name = g_strdup_printf("/cpus/cpu@%d",
239             s->soc[socket].hartid_base + cpu);
240         qemu_fdt_add_subnode(ms->fdt, cpu_name);
241 
242         if (cpu_ptr->cfg.satp_mode.supported != 0) {
243             satp_mode_max = satp_mode_max_from_map(cpu_ptr->cfg.satp_mode.map);
244             sv_name = g_strdup_printf("riscv,%s",
245                                       satp_mode_str(satp_mode_max, is_32_bit));
246             qemu_fdt_setprop_string(ms->fdt, cpu_name, "mmu-type", sv_name);
247         }
248 
249         riscv_isa_write_fdt(cpu_ptr, ms->fdt, cpu_name);
250 
251         if (cpu_ptr->cfg.ext_zicbom) {
252             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cbom-block-size",
253                                   cpu_ptr->cfg.cbom_blocksize);
254         }
255 
256         if (cpu_ptr->cfg.ext_zicboz) {
257             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cboz-block-size",
258                                   cpu_ptr->cfg.cboz_blocksize);
259         }
260 
261         if (cpu_ptr->cfg.ext_zicbop) {
262             qemu_fdt_setprop_cell(ms->fdt, cpu_name, "riscv,cbop-block-size",
263                                   cpu_ptr->cfg.cbop_blocksize);
264         }
265 
266         qemu_fdt_setprop_string(ms->fdt, cpu_name, "compatible", "riscv");
267         qemu_fdt_setprop_string(ms->fdt, cpu_name, "status", "okay");
268         qemu_fdt_setprop_cell(ms->fdt, cpu_name, "reg",
269             s->soc[socket].hartid_base + cpu);
270         qemu_fdt_setprop_string(ms->fdt, cpu_name, "device_type", "cpu");
271         riscv_socket_fdt_write_id(ms, cpu_name, socket);
272         qemu_fdt_setprop_cell(ms->fdt, cpu_name, "phandle", cpu_phandle);
273 
274         intc_phandles[cpu] = (*phandle)++;
275 
276         intc_name = g_strdup_printf("%s/interrupt-controller", cpu_name);
277         qemu_fdt_add_subnode(ms->fdt, intc_name);
278         qemu_fdt_setprop_cell(ms->fdt, intc_name, "phandle",
279             intc_phandles[cpu]);
280         qemu_fdt_setprop_string(ms->fdt, intc_name, "compatible",
281             "riscv,cpu-intc");
282         qemu_fdt_setprop(ms->fdt, intc_name, "interrupt-controller", NULL, 0);
283         qemu_fdt_setprop_cell(ms->fdt, intc_name, "#interrupt-cells", 1);
284 
285         core_name = g_strdup_printf("%s/core%d", clust_name, cpu);
286         qemu_fdt_add_subnode(ms->fdt, core_name);
287         qemu_fdt_setprop_cell(ms->fdt, core_name, "cpu", cpu_phandle);
288     }
289 }
290 
291 static void create_fdt_socket_memory(RISCVVirtState *s,
292                                      const MemMapEntry *memmap, int socket)
293 {
294     g_autofree char *mem_name = NULL;
295     uint64_t addr, size;
296     MachineState *ms = MACHINE(s);
297 
298     addr = memmap[VIRT_DRAM].base + riscv_socket_mem_offset(ms, socket);
299     size = riscv_socket_mem_size(ms, socket);
300     mem_name = g_strdup_printf("/memory@%lx", (long)addr);
301     qemu_fdt_add_subnode(ms->fdt, mem_name);
302     qemu_fdt_setprop_cells(ms->fdt, mem_name, "reg",
303         addr >> 32, addr, size >> 32, size);
304     qemu_fdt_setprop_string(ms->fdt, mem_name, "device_type", "memory");
305     riscv_socket_fdt_write_id(ms, mem_name, socket);
306 }
307 
308 static void create_fdt_socket_clint(RISCVVirtState *s,
309                                     const MemMapEntry *memmap, int socket,
310                                     uint32_t *intc_phandles)
311 {
312     int cpu;
313     g_autofree char *clint_name = NULL;
314     g_autofree uint32_t *clint_cells = NULL;
315     unsigned long clint_addr;
316     MachineState *ms = MACHINE(s);
317     static const char * const clint_compat[2] = {
318         "sifive,clint0", "riscv,clint0"
319     };
320 
321     clint_cells = g_new0(uint32_t, s->soc[socket].num_harts * 4);
322 
323     for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
324         clint_cells[cpu * 4 + 0] = cpu_to_be32(intc_phandles[cpu]);
325         clint_cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_SOFT);
326         clint_cells[cpu * 4 + 2] = cpu_to_be32(intc_phandles[cpu]);
327         clint_cells[cpu * 4 + 3] = cpu_to_be32(IRQ_M_TIMER);
328     }
329 
330     clint_addr = memmap[VIRT_CLINT].base + (memmap[VIRT_CLINT].size * socket);
331     clint_name = g_strdup_printf("/soc/clint@%lx", clint_addr);
332     qemu_fdt_add_subnode(ms->fdt, clint_name);
333     qemu_fdt_setprop_string_array(ms->fdt, clint_name, "compatible",
334                                   (char **)&clint_compat,
335                                   ARRAY_SIZE(clint_compat));
336     qemu_fdt_setprop_cells(ms->fdt, clint_name, "reg",
337         0x0, clint_addr, 0x0, memmap[VIRT_CLINT].size);
338     qemu_fdt_setprop(ms->fdt, clint_name, "interrupts-extended",
339         clint_cells, s->soc[socket].num_harts * sizeof(uint32_t) * 4);
340     riscv_socket_fdt_write_id(ms, clint_name, socket);
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     g_autofree uint32_t *aclint_mswi_cells = NULL;
352     g_autofree uint32_t *aclint_sswi_cells = NULL;
353     g_autofree uint32_t *aclint_mtimer_cells = NULL;
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 
427 static void create_fdt_socket_plic(RISCVVirtState *s,
428                                    const MemMapEntry *memmap, int socket,
429                                    uint32_t *phandle, uint32_t *intc_phandles,
430                                    uint32_t *plic_phandles)
431 {
432     int cpu;
433     g_autofree char *plic_name = NULL;
434     g_autofree uint32_t *plic_cells;
435     unsigned long plic_addr;
436     MachineState *ms = MACHINE(s);
437     static const char * const plic_compat[2] = {
438         "sifive,plic-1.0.0", "riscv,plic0"
439     };
440 
441     plic_phandles[socket] = (*phandle)++;
442     plic_addr = memmap[VIRT_PLIC].base + (memmap[VIRT_PLIC].size * socket);
443     plic_name = g_strdup_printf("/soc/plic@%lx", plic_addr);
444     qemu_fdt_add_subnode(ms->fdt, plic_name);
445     qemu_fdt_setprop_cell(ms->fdt, plic_name,
446         "#interrupt-cells", FDT_PLIC_INT_CELLS);
447     qemu_fdt_setprop_cell(ms->fdt, plic_name,
448         "#address-cells", FDT_PLIC_ADDR_CELLS);
449     qemu_fdt_setprop_string_array(ms->fdt, plic_name, "compatible",
450                                   (char **)&plic_compat,
451                                   ARRAY_SIZE(plic_compat));
452     qemu_fdt_setprop(ms->fdt, plic_name, "interrupt-controller", NULL, 0);
453 
454     if (kvm_enabled()) {
455         plic_cells = g_new0(uint32_t, s->soc[socket].num_harts * 2);
456 
457         for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
458             plic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
459             plic_cells[cpu * 2 + 1] = cpu_to_be32(IRQ_S_EXT);
460         }
461 
462         qemu_fdt_setprop(ms->fdt, plic_name, "interrupts-extended",
463                          plic_cells,
464                          s->soc[socket].num_harts * sizeof(uint32_t) * 2);
465    } else {
466         plic_cells = g_new0(uint32_t, s->soc[socket].num_harts * 4);
467 
468         for (cpu = 0; cpu < s->soc[socket].num_harts; cpu++) {
469             plic_cells[cpu * 4 + 0] = cpu_to_be32(intc_phandles[cpu]);
470             plic_cells[cpu * 4 + 1] = cpu_to_be32(IRQ_M_EXT);
471             plic_cells[cpu * 4 + 2] = cpu_to_be32(intc_phandles[cpu]);
472             plic_cells[cpu * 4 + 3] = cpu_to_be32(IRQ_S_EXT);
473         }
474 
475         qemu_fdt_setprop(ms->fdt, plic_name, "interrupts-extended",
476                          plic_cells,
477                          s->soc[socket].num_harts * sizeof(uint32_t) * 4);
478     }
479 
480     qemu_fdt_setprop_cells(ms->fdt, plic_name, "reg",
481         0x0, plic_addr, 0x0, memmap[VIRT_PLIC].size);
482     qemu_fdt_setprop_cell(ms->fdt, plic_name, "riscv,ndev",
483                           VIRT_IRQCHIP_NUM_SOURCES - 1);
484     riscv_socket_fdt_write_id(ms, plic_name, socket);
485     qemu_fdt_setprop_cell(ms->fdt, plic_name, "phandle",
486         plic_phandles[socket]);
487 
488     if (!socket) {
489         platform_bus_add_all_fdt_nodes(ms->fdt, plic_name,
490                                        memmap[VIRT_PLATFORM_BUS].base,
491                                        memmap[VIRT_PLATFORM_BUS].size,
492                                        VIRT_PLATFORM_BUS_IRQ);
493     }
494 }
495 
496 uint32_t imsic_num_bits(uint32_t count)
497 {
498     uint32_t ret = 0;
499 
500     while (BIT(ret) < count) {
501         ret++;
502     }
503 
504     return ret;
505 }
506 
507 static void create_fdt_one_imsic(RISCVVirtState *s, hwaddr base_addr,
508                                  uint32_t *intc_phandles, uint32_t msi_phandle,
509                                  bool m_mode, uint32_t imsic_guest_bits)
510 {
511     int cpu, socket;
512     g_autofree char *imsic_name = NULL;
513     MachineState *ms = MACHINE(s);
514     int socket_count = riscv_socket_count(ms);
515     uint32_t imsic_max_hart_per_socket, imsic_addr, imsic_size;
516     g_autofree uint32_t *imsic_cells = NULL;
517     g_autofree uint32_t *imsic_regs = NULL;
518     static const char * const imsic_compat[2] = {
519         "qemu,imsics", "riscv,imsics"
520     };
521 
522     imsic_cells = g_new0(uint32_t, ms->smp.cpus * 2);
523     imsic_regs = g_new0(uint32_t, socket_count * 4);
524 
525     for (cpu = 0; cpu < ms->smp.cpus; cpu++) {
526         imsic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
527         imsic_cells[cpu * 2 + 1] = cpu_to_be32(m_mode ? IRQ_M_EXT : IRQ_S_EXT);
528     }
529 
530     imsic_max_hart_per_socket = 0;
531     for (socket = 0; socket < socket_count; socket++) {
532         imsic_addr = base_addr + socket * VIRT_IMSIC_GROUP_MAX_SIZE;
533         imsic_size = IMSIC_HART_SIZE(imsic_guest_bits) *
534                      s->soc[socket].num_harts;
535         imsic_regs[socket * 4 + 0] = 0;
536         imsic_regs[socket * 4 + 1] = cpu_to_be32(imsic_addr);
537         imsic_regs[socket * 4 + 2] = 0;
538         imsic_regs[socket * 4 + 3] = cpu_to_be32(imsic_size);
539         if (imsic_max_hart_per_socket < s->soc[socket].num_harts) {
540             imsic_max_hart_per_socket = s->soc[socket].num_harts;
541         }
542     }
543 
544     imsic_name = g_strdup_printf("/soc/interrupt-controller@%lx",
545                                  (unsigned long)base_addr);
546     qemu_fdt_add_subnode(ms->fdt, imsic_name);
547     qemu_fdt_setprop_string_array(ms->fdt, imsic_name, "compatible",
548                                   (char **)&imsic_compat,
549                                   ARRAY_SIZE(imsic_compat));
550 
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_cell(ms->fdt, imsic_name, "#msi-cells", 0);
556     qemu_fdt_setprop(ms->fdt, imsic_name, "interrupts-extended",
557                      imsic_cells, ms->smp.cpus * sizeof(uint32_t) * 2);
558     qemu_fdt_setprop(ms->fdt, imsic_name, "reg", imsic_regs,
559                      socket_count * sizeof(uint32_t) * 4);
560     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,num-ids",
561                      VIRT_IRQCHIP_NUM_MSIS);
562 
563     if (imsic_guest_bits) {
564         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,guest-index-bits",
565                               imsic_guest_bits);
566     }
567 
568     if (socket_count > 1) {
569         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,hart-index-bits",
570                               imsic_num_bits(imsic_max_hart_per_socket));
571         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-bits",
572                               imsic_num_bits(socket_count));
573         qemu_fdt_setprop_cell(ms->fdt, imsic_name, "riscv,group-index-shift",
574                               IMSIC_MMIO_GROUP_MIN_SHIFT);
575     }
576     qemu_fdt_setprop_cell(ms->fdt, imsic_name, "phandle", msi_phandle);
577 }
578 
579 static void create_fdt_imsic(RISCVVirtState *s, const MemMapEntry *memmap,
580                              uint32_t *phandle, uint32_t *intc_phandles,
581                              uint32_t *msi_m_phandle, uint32_t *msi_s_phandle)
582 {
583     *msi_m_phandle = (*phandle)++;
584     *msi_s_phandle = (*phandle)++;
585 
586     if (!kvm_enabled()) {
587         /* M-level IMSIC node */
588         create_fdt_one_imsic(s, memmap[VIRT_IMSIC_M].base, intc_phandles,
589                              *msi_m_phandle, true, 0);
590     }
591 
592     /* S-level IMSIC node */
593     create_fdt_one_imsic(s, memmap[VIRT_IMSIC_S].base, intc_phandles,
594                          *msi_s_phandle, false,
595                          imsic_num_bits(s->aia_guests + 1));
596 
597 }
598 
599 /* Caller must free string after use */
600 static char *fdt_get_aplic_nodename(unsigned long aplic_addr)
601 {
602     return g_strdup_printf("/soc/interrupt-controller@%lx", aplic_addr);
603 }
604 
605 static void create_fdt_one_aplic(RISCVVirtState *s, int socket,
606                                  unsigned long aplic_addr, uint32_t aplic_size,
607                                  uint32_t msi_phandle,
608                                  uint32_t *intc_phandles,
609                                  uint32_t aplic_phandle,
610                                  uint32_t aplic_child_phandle,
611                                  bool m_mode, int num_harts)
612 {
613     int cpu;
614     g_autofree char *aplic_name = fdt_get_aplic_nodename(aplic_addr);
615     g_autofree uint32_t *aplic_cells = g_new0(uint32_t, num_harts * 2);
616     MachineState *ms = MACHINE(s);
617     static const char * const aplic_compat[2] = {
618         "qemu,aplic", "riscv,aplic"
619     };
620 
621     for (cpu = 0; cpu < num_harts; cpu++) {
622         aplic_cells[cpu * 2 + 0] = cpu_to_be32(intc_phandles[cpu]);
623         aplic_cells[cpu * 2 + 1] = cpu_to_be32(m_mode ? IRQ_M_EXT : IRQ_S_EXT);
624     }
625 
626     qemu_fdt_add_subnode(ms->fdt, aplic_name);
627     qemu_fdt_setprop_string_array(ms->fdt, aplic_name, "compatible",
628                                   (char **)&aplic_compat,
629                                   ARRAY_SIZE(aplic_compat));
630     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "#address-cells",
631                           FDT_APLIC_ADDR_CELLS);
632     qemu_fdt_setprop_cell(ms->fdt, aplic_name,
633                           "#interrupt-cells", FDT_APLIC_INT_CELLS);
634     qemu_fdt_setprop(ms->fdt, aplic_name, "interrupt-controller", NULL, 0);
635 
636     if (s->aia_type == VIRT_AIA_TYPE_APLIC) {
637         qemu_fdt_setprop(ms->fdt, aplic_name, "interrupts-extended",
638                          aplic_cells, num_harts * sizeof(uint32_t) * 2);
639     } else {
640         qemu_fdt_setprop_cell(ms->fdt, aplic_name, "msi-parent", msi_phandle);
641     }
642 
643     qemu_fdt_setprop_cells(ms->fdt, aplic_name, "reg",
644                            0x0, aplic_addr, 0x0, aplic_size);
645     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "riscv,num-sources",
646                           VIRT_IRQCHIP_NUM_SOURCES);
647 
648     if (aplic_child_phandle) {
649         qemu_fdt_setprop_cell(ms->fdt, aplic_name, "riscv,children",
650                               aplic_child_phandle);
651         qemu_fdt_setprop_cells(ms->fdt, aplic_name, "riscv,delegation",
652                                aplic_child_phandle, 0x1,
653                                VIRT_IRQCHIP_NUM_SOURCES);
654     }
655 
656     riscv_socket_fdt_write_id(ms, aplic_name, socket);
657     qemu_fdt_setprop_cell(ms->fdt, aplic_name, "phandle", aplic_phandle);
658 }
659 
660 static void create_fdt_socket_aplic(RISCVVirtState *s,
661                                     const MemMapEntry *memmap, int socket,
662                                     uint32_t msi_m_phandle,
663                                     uint32_t msi_s_phandle,
664                                     uint32_t *phandle,
665                                     uint32_t *intc_phandles,
666                                     uint32_t *aplic_phandles,
667                                     int num_harts)
668 {
669     unsigned long aplic_addr;
670     MachineState *ms = MACHINE(s);
671     uint32_t aplic_m_phandle, aplic_s_phandle;
672 
673     aplic_m_phandle = (*phandle)++;
674     aplic_s_phandle = (*phandle)++;
675 
676     if (!kvm_enabled()) {
677         /* M-level APLIC node */
678         aplic_addr = memmap[VIRT_APLIC_M].base +
679                      (memmap[VIRT_APLIC_M].size * socket);
680         create_fdt_one_aplic(s, socket, aplic_addr, memmap[VIRT_APLIC_M].size,
681                              msi_m_phandle, intc_phandles,
682                              aplic_m_phandle, aplic_s_phandle,
683                              true, num_harts);
684     }
685 
686     /* S-level APLIC node */
687     aplic_addr = memmap[VIRT_APLIC_S].base +
688                  (memmap[VIRT_APLIC_S].size * socket);
689     create_fdt_one_aplic(s, socket, aplic_addr, memmap[VIRT_APLIC_S].size,
690                          msi_s_phandle, intc_phandles,
691                          aplic_s_phandle, 0,
692                          false, num_harts);
693 
694     if (!socket) {
695         g_autofree char *aplic_name = fdt_get_aplic_nodename(aplic_addr);
696         platform_bus_add_all_fdt_nodes(ms->fdt, aplic_name,
697                                        memmap[VIRT_PLATFORM_BUS].base,
698                                        memmap[VIRT_PLATFORM_BUS].size,
699                                        VIRT_PLATFORM_BUS_IRQ);
700     }
701 
702     aplic_phandles[socket] = aplic_s_phandle;
703 }
704 
705 static void create_fdt_pmu(RISCVVirtState *s)
706 {
707     g_autofree char *pmu_name = g_strdup_printf("/pmu");
708     MachineState *ms = MACHINE(s);
709     RISCVCPU hart = s->soc[0].harts[0];
710 
711     qemu_fdt_add_subnode(ms->fdt, pmu_name);
712     qemu_fdt_setprop_string(ms->fdt, pmu_name, "compatible", "riscv,pmu");
713     riscv_pmu_generate_fdt_node(ms->fdt, hart.pmu_avail_ctrs, pmu_name);
714 }
715 
716 static void create_fdt_sockets(RISCVVirtState *s, const MemMapEntry *memmap,
717                                uint32_t *phandle,
718                                uint32_t *irq_mmio_phandle,
719                                uint32_t *irq_pcie_phandle,
720                                uint32_t *irq_virtio_phandle,
721                                uint32_t *msi_pcie_phandle)
722 {
723     int socket, phandle_pos;
724     MachineState *ms = MACHINE(s);
725     uint32_t msi_m_phandle = 0, msi_s_phandle = 0;
726     uint32_t xplic_phandles[MAX_NODES];
727     g_autofree uint32_t *intc_phandles = NULL;
728     int socket_count = riscv_socket_count(ms);
729 
730     qemu_fdt_add_subnode(ms->fdt, "/cpus");
731     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "timebase-frequency",
732                           kvm_enabled() ?
733                           kvm_riscv_get_timebase_frequency(first_cpu) :
734                           RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ);
735     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
736     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", 0x1);
737     qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
738 
739     intc_phandles = g_new0(uint32_t, ms->smp.cpus);
740 
741     phandle_pos = ms->smp.cpus;
742     for (socket = (socket_count - 1); socket >= 0; socket--) {
743         g_autofree char *clust_name = NULL;
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         if (virt_aclint_allowed() && s->have_aclint) {
755             create_fdt_socket_aclint(s, memmap, socket,
756                                      &intc_phandles[phandle_pos]);
757         } else if (tcg_enabled()) {
758             create_fdt_socket_clint(s, memmap, socket,
759                                     &intc_phandles[phandle_pos]);
760         }
761     }
762 
763     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
764         create_fdt_imsic(s, memmap, phandle, intc_phandles,
765             &msi_m_phandle, &msi_s_phandle);
766         *msi_pcie_phandle = msi_s_phandle;
767     }
768 
769     /* KVM AIA only has one APLIC instance */
770     if (kvm_enabled() && virt_use_kvm_aia(s)) {
771         create_fdt_socket_aplic(s, memmap, 0,
772                                 msi_m_phandle, msi_s_phandle, phandle,
773                                 &intc_phandles[0], xplic_phandles,
774                                 ms->smp.cpus);
775     } else {
776         phandle_pos = ms->smp.cpus;
777         for (socket = (socket_count - 1); socket >= 0; socket--) {
778             phandle_pos -= s->soc[socket].num_harts;
779 
780             if (s->aia_type == VIRT_AIA_TYPE_NONE) {
781                 create_fdt_socket_plic(s, memmap, socket, phandle,
782                                        &intc_phandles[phandle_pos],
783                                        xplic_phandles);
784             } else {
785                 create_fdt_socket_aplic(s, memmap, socket,
786                                         msi_m_phandle, msi_s_phandle, phandle,
787                                         &intc_phandles[phandle_pos],
788                                         xplic_phandles,
789                                         s->soc[socket].num_harts);
790             }
791         }
792     }
793 
794     if (kvm_enabled() && virt_use_kvm_aia(s)) {
795         *irq_mmio_phandle = xplic_phandles[0];
796         *irq_virtio_phandle = xplic_phandles[0];
797         *irq_pcie_phandle = xplic_phandles[0];
798     } else {
799         for (socket = 0; socket < socket_count; socket++) {
800             if (socket == 0) {
801                 *irq_mmio_phandle = xplic_phandles[socket];
802                 *irq_virtio_phandle = xplic_phandles[socket];
803                 *irq_pcie_phandle = xplic_phandles[socket];
804             }
805             if (socket == 1) {
806                 *irq_virtio_phandle = xplic_phandles[socket];
807                 *irq_pcie_phandle = xplic_phandles[socket];
808             }
809             if (socket == 2) {
810                 *irq_pcie_phandle = xplic_phandles[socket];
811             }
812         }
813     }
814 
815     riscv_socket_fdt_write_distance_matrix(ms);
816 }
817 
818 static void create_fdt_virtio(RISCVVirtState *s, const MemMapEntry *memmap,
819                               uint32_t irq_virtio_phandle)
820 {
821     int i;
822     MachineState *ms = MACHINE(s);
823 
824     for (i = 0; i < VIRTIO_COUNT; i++) {
825         g_autofree char *name =  g_strdup_printf("/soc/virtio_mmio@%lx",
826             (long)(memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size));
827 
828         qemu_fdt_add_subnode(ms->fdt, name);
829         qemu_fdt_setprop_string(ms->fdt, name, "compatible", "virtio,mmio");
830         qemu_fdt_setprop_cells(ms->fdt, name, "reg",
831             0x0, memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size,
832             0x0, memmap[VIRT_VIRTIO].size);
833         qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent",
834             irq_virtio_phandle);
835         if (s->aia_type == VIRT_AIA_TYPE_NONE) {
836             qemu_fdt_setprop_cell(ms->fdt, name, "interrupts",
837                                   VIRTIO_IRQ + i);
838         } else {
839             qemu_fdt_setprop_cells(ms->fdt, name, "interrupts",
840                                    VIRTIO_IRQ + i, 0x4);
841         }
842     }
843 }
844 
845 static void create_fdt_pcie(RISCVVirtState *s, const MemMapEntry *memmap,
846                             uint32_t irq_pcie_phandle,
847                             uint32_t msi_pcie_phandle)
848 {
849     g_autofree char *name = NULL;
850     MachineState *ms = MACHINE(s);
851 
852     name = g_strdup_printf("/soc/pci@%lx",
853         (long) memmap[VIRT_PCIE_ECAM].base);
854     qemu_fdt_setprop_cell(ms->fdt, name, "#address-cells",
855         FDT_PCI_ADDR_CELLS);
856     qemu_fdt_setprop_cell(ms->fdt, name, "#interrupt-cells",
857         FDT_PCI_INT_CELLS);
858     qemu_fdt_setprop_cell(ms->fdt, name, "#size-cells", 0x2);
859     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
860         "pci-host-ecam-generic");
861     qemu_fdt_setprop_string(ms->fdt, name, "device_type", "pci");
862     qemu_fdt_setprop_cell(ms->fdt, name, "linux,pci-domain", 0);
863     qemu_fdt_setprop_cells(ms->fdt, name, "bus-range", 0,
864         memmap[VIRT_PCIE_ECAM].size / PCIE_MMCFG_SIZE_MIN - 1);
865     qemu_fdt_setprop(ms->fdt, name, "dma-coherent", NULL, 0);
866     if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
867         qemu_fdt_setprop_cell(ms->fdt, name, "msi-parent", msi_pcie_phandle);
868     }
869     qemu_fdt_setprop_cells(ms->fdt, name, "reg", 0,
870         memmap[VIRT_PCIE_ECAM].base, 0, memmap[VIRT_PCIE_ECAM].size);
871     qemu_fdt_setprop_sized_cells(ms->fdt, name, "ranges",
872         1, FDT_PCI_RANGE_IOPORT, 2, 0,
873         2, memmap[VIRT_PCIE_PIO].base, 2, memmap[VIRT_PCIE_PIO].size,
874         1, FDT_PCI_RANGE_MMIO,
875         2, memmap[VIRT_PCIE_MMIO].base,
876         2, memmap[VIRT_PCIE_MMIO].base, 2, memmap[VIRT_PCIE_MMIO].size,
877         1, FDT_PCI_RANGE_MMIO_64BIT,
878         2, virt_high_pcie_memmap.base,
879         2, virt_high_pcie_memmap.base, 2, virt_high_pcie_memmap.size);
880 
881     create_pcie_irq_map(s, ms->fdt, name, irq_pcie_phandle);
882 }
883 
884 static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap,
885                              uint32_t *phandle)
886 {
887     char *name;
888     uint32_t test_phandle;
889     MachineState *ms = MACHINE(s);
890 
891     test_phandle = (*phandle)++;
892     name = g_strdup_printf("/soc/test@%lx",
893         (long)memmap[VIRT_TEST].base);
894     qemu_fdt_add_subnode(ms->fdt, name);
895     {
896         static const char * const compat[3] = {
897             "sifive,test1", "sifive,test0", "syscon"
898         };
899         qemu_fdt_setprop_string_array(ms->fdt, name, "compatible",
900                                       (char **)&compat, ARRAY_SIZE(compat));
901     }
902     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
903         0x0, memmap[VIRT_TEST].base, 0x0, memmap[VIRT_TEST].size);
904     qemu_fdt_setprop_cell(ms->fdt, name, "phandle", test_phandle);
905     test_phandle = qemu_fdt_get_phandle(ms->fdt, name);
906     g_free(name);
907 
908     name = g_strdup_printf("/reboot");
909     qemu_fdt_add_subnode(ms->fdt, name);
910     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "syscon-reboot");
911     qemu_fdt_setprop_cell(ms->fdt, name, "regmap", test_phandle);
912     qemu_fdt_setprop_cell(ms->fdt, name, "offset", 0x0);
913     qemu_fdt_setprop_cell(ms->fdt, name, "value", FINISHER_RESET);
914     g_free(name);
915 
916     name = g_strdup_printf("/poweroff");
917     qemu_fdt_add_subnode(ms->fdt, name);
918     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "syscon-poweroff");
919     qemu_fdt_setprop_cell(ms->fdt, name, "regmap", test_phandle);
920     qemu_fdt_setprop_cell(ms->fdt, name, "offset", 0x0);
921     qemu_fdt_setprop_cell(ms->fdt, name, "value", FINISHER_PASS);
922     g_free(name);
923 }
924 
925 static void create_fdt_uart(RISCVVirtState *s, const MemMapEntry *memmap,
926                             uint32_t irq_mmio_phandle)
927 {
928     g_autofree char *name = NULL;
929     MachineState *ms = MACHINE(s);
930 
931     name = g_strdup_printf("/soc/serial@%lx", (long)memmap[VIRT_UART0].base);
932     qemu_fdt_add_subnode(ms->fdt, name);
933     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "ns16550a");
934     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
935         0x0, memmap[VIRT_UART0].base,
936         0x0, memmap[VIRT_UART0].size);
937     qemu_fdt_setprop_cell(ms->fdt, name, "clock-frequency", 3686400);
938     qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent", irq_mmio_phandle);
939     if (s->aia_type == VIRT_AIA_TYPE_NONE) {
940         qemu_fdt_setprop_cell(ms->fdt, name, "interrupts", UART0_IRQ);
941     } else {
942         qemu_fdt_setprop_cells(ms->fdt, name, "interrupts", UART0_IRQ, 0x4);
943     }
944 
945     qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", name);
946 }
947 
948 static void create_fdt_rtc(RISCVVirtState *s, const MemMapEntry *memmap,
949                            uint32_t irq_mmio_phandle)
950 {
951     g_autofree char *name = NULL;
952     MachineState *ms = MACHINE(s);
953 
954     name = g_strdup_printf("/soc/rtc@%lx", (long)memmap[VIRT_RTC].base);
955     qemu_fdt_add_subnode(ms->fdt, name);
956     qemu_fdt_setprop_string(ms->fdt, name, "compatible",
957         "google,goldfish-rtc");
958     qemu_fdt_setprop_cells(ms->fdt, name, "reg",
959         0x0, memmap[VIRT_RTC].base, 0x0, memmap[VIRT_RTC].size);
960     qemu_fdt_setprop_cell(ms->fdt, name, "interrupt-parent",
961         irq_mmio_phandle);
962     if (s->aia_type == VIRT_AIA_TYPE_NONE) {
963         qemu_fdt_setprop_cell(ms->fdt, name, "interrupts", RTC_IRQ);
964     } else {
965         qemu_fdt_setprop_cells(ms->fdt, name, "interrupts", RTC_IRQ, 0x4);
966     }
967 }
968 
969 static void create_fdt_flash(RISCVVirtState *s, const MemMapEntry *memmap)
970 {
971     MachineState *ms = MACHINE(s);
972     hwaddr flashsize = virt_memmap[VIRT_FLASH].size / 2;
973     hwaddr flashbase = virt_memmap[VIRT_FLASH].base;
974     g_autofree char *name = g_strdup_printf("/flash@%" PRIx64, flashbase);
975 
976     qemu_fdt_add_subnode(ms->fdt, name);
977     qemu_fdt_setprop_string(ms->fdt, name, "compatible", "cfi-flash");
978     qemu_fdt_setprop_sized_cells(ms->fdt, name, "reg",
979                                  2, flashbase, 2, flashsize,
980                                  2, flashbase + flashsize, 2, flashsize);
981     qemu_fdt_setprop_cell(ms->fdt, name, "bank-width", 4);
982 }
983 
984 static void create_fdt_fw_cfg(RISCVVirtState *s, const MemMapEntry *memmap)
985 {
986     MachineState *ms = MACHINE(s);
987     hwaddr base = memmap[VIRT_FW_CFG].base;
988     hwaddr size = memmap[VIRT_FW_CFG].size;
989     g_autofree char *nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
990 
991     qemu_fdt_add_subnode(ms->fdt, nodename);
992     qemu_fdt_setprop_string(ms->fdt, nodename,
993                             "compatible", "qemu,fw-cfg-mmio");
994     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
995                                  2, base, 2, size);
996     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
997 }
998 
999 static void create_fdt_virtio_iommu(RISCVVirtState *s, uint16_t bdf)
1000 {
1001     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1002     void *fdt = MACHINE(s)->fdt;
1003     uint32_t iommu_phandle;
1004     g_autofree char *iommu_node = NULL;
1005     g_autofree char *pci_node = NULL;
1006 
1007     pci_node = g_strdup_printf("/soc/pci@%lx",
1008                                (long) virt_memmap[VIRT_PCIE_ECAM].base);
1009     iommu_node = g_strdup_printf("%s/virtio_iommu@%x,%x", pci_node,
1010                                  PCI_SLOT(bdf), PCI_FUNC(bdf));
1011     iommu_phandle = qemu_fdt_alloc_phandle(fdt);
1012 
1013     qemu_fdt_add_subnode(fdt, iommu_node);
1014 
1015     qemu_fdt_setprop(fdt, iommu_node, "compatible", compat, sizeof(compat));
1016     qemu_fdt_setprop_sized_cells(fdt, iommu_node, "reg",
1017                                  1, bdf << 8, 1, 0, 1, 0,
1018                                  1, 0, 1, 0);
1019     qemu_fdt_setprop_cell(fdt, iommu_node, "#iommu-cells", 1);
1020     qemu_fdt_setprop_cell(fdt, iommu_node, "phandle", iommu_phandle);
1021 
1022     qemu_fdt_setprop_cells(fdt, pci_node, "iommu-map",
1023                            0, iommu_phandle, 0, bdf,
1024                            bdf + 1, iommu_phandle, bdf + 1, 0xffff - bdf);
1025 }
1026 
1027 static void finalize_fdt(RISCVVirtState *s)
1028 {
1029     uint32_t phandle = 1, irq_mmio_phandle = 1, msi_pcie_phandle = 1;
1030     uint32_t irq_pcie_phandle = 1, irq_virtio_phandle = 1;
1031 
1032     create_fdt_sockets(s, virt_memmap, &phandle, &irq_mmio_phandle,
1033                        &irq_pcie_phandle, &irq_virtio_phandle,
1034                        &msi_pcie_phandle);
1035 
1036     create_fdt_virtio(s, virt_memmap, irq_virtio_phandle);
1037 
1038     create_fdt_pcie(s, virt_memmap, irq_pcie_phandle, msi_pcie_phandle);
1039 
1040     create_fdt_reset(s, virt_memmap, &phandle);
1041 
1042     create_fdt_uart(s, virt_memmap, irq_mmio_phandle);
1043 
1044     create_fdt_rtc(s, virt_memmap, irq_mmio_phandle);
1045 }
1046 
1047 static void create_fdt(RISCVVirtState *s, const MemMapEntry *memmap)
1048 {
1049     MachineState *ms = MACHINE(s);
1050     uint8_t rng_seed[32];
1051     g_autofree char *name = NULL;
1052 
1053     ms->fdt = create_device_tree(&s->fdt_size);
1054     if (!ms->fdt) {
1055         error_report("create_device_tree() failed");
1056         exit(1);
1057     }
1058 
1059     qemu_fdt_setprop_string(ms->fdt, "/", "model", "riscv-virtio,qemu");
1060     qemu_fdt_setprop_string(ms->fdt, "/", "compatible", "riscv-virtio");
1061     qemu_fdt_setprop_cell(ms->fdt, "/", "#size-cells", 0x2);
1062     qemu_fdt_setprop_cell(ms->fdt, "/", "#address-cells", 0x2);
1063 
1064     qemu_fdt_add_subnode(ms->fdt, "/soc");
1065     qemu_fdt_setprop(ms->fdt, "/soc", "ranges", NULL, 0);
1066     qemu_fdt_setprop_string(ms->fdt, "/soc", "compatible", "simple-bus");
1067     qemu_fdt_setprop_cell(ms->fdt, "/soc", "#size-cells", 0x2);
1068     qemu_fdt_setprop_cell(ms->fdt, "/soc", "#address-cells", 0x2);
1069 
1070     /*
1071      * The "/soc/pci@..." node is needed for PCIE hotplugs
1072      * that might happen before finalize_fdt().
1073      */
1074     name = g_strdup_printf("/soc/pci@%lx", (long) memmap[VIRT_PCIE_ECAM].base);
1075     qemu_fdt_add_subnode(ms->fdt, name);
1076 
1077     qemu_fdt_add_subnode(ms->fdt, "/chosen");
1078 
1079     /* Pass seed to RNG */
1080     qemu_guest_getrandom_nofail(rng_seed, sizeof(rng_seed));
1081     qemu_fdt_setprop(ms->fdt, "/chosen", "rng-seed",
1082                      rng_seed, sizeof(rng_seed));
1083 
1084     create_fdt_flash(s, memmap);
1085     create_fdt_fw_cfg(s, memmap);
1086     create_fdt_pmu(s);
1087 }
1088 
1089 static inline DeviceState *gpex_pcie_init(MemoryRegion *sys_mem,
1090                                           DeviceState *irqchip,
1091                                           RISCVVirtState *s)
1092 {
1093     DeviceState *dev;
1094     MemoryRegion *ecam_alias, *ecam_reg;
1095     MemoryRegion *mmio_alias, *high_mmio_alias, *mmio_reg;
1096     hwaddr ecam_base = s->memmap[VIRT_PCIE_ECAM].base;
1097     hwaddr ecam_size = s->memmap[VIRT_PCIE_ECAM].size;
1098     hwaddr mmio_base = s->memmap[VIRT_PCIE_MMIO].base;
1099     hwaddr mmio_size = s->memmap[VIRT_PCIE_MMIO].size;
1100     hwaddr high_mmio_base = virt_high_pcie_memmap.base;
1101     hwaddr high_mmio_size = virt_high_pcie_memmap.size;
1102     hwaddr pio_base = s->memmap[VIRT_PCIE_PIO].base;
1103     hwaddr pio_size = s->memmap[VIRT_PCIE_PIO].size;
1104     qemu_irq irq;
1105     int i;
1106 
1107     dev = qdev_new(TYPE_GPEX_HOST);
1108 
1109     /* Set GPEX object properties for the virt machine */
1110     object_property_set_uint(OBJECT(GPEX_HOST(dev)), PCI_HOST_ECAM_BASE,
1111                             ecam_base, NULL);
1112     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_ECAM_SIZE,
1113                             ecam_size, NULL);
1114     object_property_set_uint(OBJECT(GPEX_HOST(dev)),
1115                              PCI_HOST_BELOW_4G_MMIO_BASE,
1116                              mmio_base, NULL);
1117     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_BELOW_4G_MMIO_SIZE,
1118                             mmio_size, NULL);
1119     object_property_set_uint(OBJECT(GPEX_HOST(dev)),
1120                              PCI_HOST_ABOVE_4G_MMIO_BASE,
1121                              high_mmio_base, NULL);
1122     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_ABOVE_4G_MMIO_SIZE,
1123                             high_mmio_size, NULL);
1124     object_property_set_uint(OBJECT(GPEX_HOST(dev)), PCI_HOST_PIO_BASE,
1125                             pio_base, NULL);
1126     object_property_set_int(OBJECT(GPEX_HOST(dev)), PCI_HOST_PIO_SIZE,
1127                             pio_size, NULL);
1128 
1129     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1130 
1131     ecam_alias = g_new0(MemoryRegion, 1);
1132     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1133     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1134                              ecam_reg, 0, ecam_size);
1135     memory_region_add_subregion(get_system_memory(), ecam_base, ecam_alias);
1136 
1137     mmio_alias = g_new0(MemoryRegion, 1);
1138     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1139     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1140                              mmio_reg, mmio_base, mmio_size);
1141     memory_region_add_subregion(get_system_memory(), mmio_base, mmio_alias);
1142 
1143     /* Map high MMIO space */
1144     high_mmio_alias = g_new0(MemoryRegion, 1);
1145     memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1146                              mmio_reg, high_mmio_base, high_mmio_size);
1147     memory_region_add_subregion(get_system_memory(), high_mmio_base,
1148                                 high_mmio_alias);
1149 
1150     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, pio_base);
1151 
1152     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1153         irq = qdev_get_gpio_in(irqchip, PCIE_IRQ + i);
1154 
1155         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, irq);
1156         gpex_set_irq_num(GPEX_HOST(dev), i, PCIE_IRQ + i);
1157     }
1158 
1159     GPEX_HOST(dev)->gpex_cfg.bus = PCI_HOST_BRIDGE(GPEX_HOST(dev))->bus;
1160     return dev;
1161 }
1162 
1163 static FWCfgState *create_fw_cfg(const MachineState *ms)
1164 {
1165     hwaddr base = virt_memmap[VIRT_FW_CFG].base;
1166     FWCfgState *fw_cfg;
1167 
1168     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16,
1169                                   &address_space_memory);
1170     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1171 
1172     return fw_cfg;
1173 }
1174 
1175 static DeviceState *virt_create_plic(const MemMapEntry *memmap, int socket,
1176                                      int base_hartid, int hart_count)
1177 {
1178     DeviceState *ret;
1179     g_autofree char *plic_hart_config = NULL;
1180 
1181     /* Per-socket PLIC hart topology configuration string */
1182     plic_hart_config = riscv_plic_hart_config_string(hart_count);
1183 
1184     /* Per-socket PLIC */
1185     ret = sifive_plic_create(
1186             memmap[VIRT_PLIC].base + socket * memmap[VIRT_PLIC].size,
1187             plic_hart_config, hart_count, base_hartid,
1188             VIRT_IRQCHIP_NUM_SOURCES,
1189             ((1U << VIRT_IRQCHIP_NUM_PRIO_BITS) - 1),
1190             VIRT_PLIC_PRIORITY_BASE,
1191             VIRT_PLIC_PENDING_BASE,
1192             VIRT_PLIC_ENABLE_BASE,
1193             VIRT_PLIC_ENABLE_STRIDE,
1194             VIRT_PLIC_CONTEXT_BASE,
1195             VIRT_PLIC_CONTEXT_STRIDE,
1196             memmap[VIRT_PLIC].size);
1197 
1198     return ret;
1199 }
1200 
1201 static DeviceState *virt_create_aia(RISCVVirtAIAType aia_type, int aia_guests,
1202                                     const MemMapEntry *memmap, int socket,
1203                                     int base_hartid, int hart_count)
1204 {
1205     int i;
1206     hwaddr addr;
1207     uint32_t guest_bits;
1208     DeviceState *aplic_s = NULL;
1209     DeviceState *aplic_m = NULL;
1210     bool msimode = aia_type == VIRT_AIA_TYPE_APLIC_IMSIC;
1211 
1212     if (msimode) {
1213         if (!kvm_enabled()) {
1214             /* Per-socket M-level IMSICs */
1215             addr = memmap[VIRT_IMSIC_M].base +
1216                    socket * VIRT_IMSIC_GROUP_MAX_SIZE;
1217             for (i = 0; i < hart_count; i++) {
1218                 riscv_imsic_create(addr + i * IMSIC_HART_SIZE(0),
1219                                    base_hartid + i, true, 1,
1220                                    VIRT_IRQCHIP_NUM_MSIS);
1221             }
1222         }
1223 
1224         /* Per-socket S-level IMSICs */
1225         guest_bits = imsic_num_bits(aia_guests + 1);
1226         addr = memmap[VIRT_IMSIC_S].base + socket * VIRT_IMSIC_GROUP_MAX_SIZE;
1227         for (i = 0; i < hart_count; i++) {
1228             riscv_imsic_create(addr + i * IMSIC_HART_SIZE(guest_bits),
1229                                base_hartid + i, false, 1 + aia_guests,
1230                                VIRT_IRQCHIP_NUM_MSIS);
1231         }
1232     }
1233 
1234     if (!kvm_enabled()) {
1235         /* Per-socket M-level APLIC */
1236         aplic_m = riscv_aplic_create(memmap[VIRT_APLIC_M].base +
1237                                      socket * memmap[VIRT_APLIC_M].size,
1238                                      memmap[VIRT_APLIC_M].size,
1239                                      (msimode) ? 0 : base_hartid,
1240                                      (msimode) ? 0 : hart_count,
1241                                      VIRT_IRQCHIP_NUM_SOURCES,
1242                                      VIRT_IRQCHIP_NUM_PRIO_BITS,
1243                                      msimode, true, NULL);
1244     }
1245 
1246     /* Per-socket S-level APLIC */
1247     aplic_s = riscv_aplic_create(memmap[VIRT_APLIC_S].base +
1248                                  socket * memmap[VIRT_APLIC_S].size,
1249                                  memmap[VIRT_APLIC_S].size,
1250                                  (msimode) ? 0 : base_hartid,
1251                                  (msimode) ? 0 : hart_count,
1252                                  VIRT_IRQCHIP_NUM_SOURCES,
1253                                  VIRT_IRQCHIP_NUM_PRIO_BITS,
1254                                  msimode, false, aplic_m);
1255 
1256     return kvm_enabled() ? aplic_s : aplic_m;
1257 }
1258 
1259 static void create_platform_bus(RISCVVirtState *s, DeviceState *irqchip)
1260 {
1261     DeviceState *dev;
1262     SysBusDevice *sysbus;
1263     const MemMapEntry *memmap = virt_memmap;
1264     int i;
1265     MemoryRegion *sysmem = get_system_memory();
1266 
1267     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1268     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1269     qdev_prop_set_uint32(dev, "num_irqs", VIRT_PLATFORM_BUS_NUM_IRQS);
1270     qdev_prop_set_uint32(dev, "mmio_size", memmap[VIRT_PLATFORM_BUS].size);
1271     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1272     s->platform_bus_dev = dev;
1273 
1274     sysbus = SYS_BUS_DEVICE(dev);
1275     for (i = 0; i < VIRT_PLATFORM_BUS_NUM_IRQS; i++) {
1276         int irq = VIRT_PLATFORM_BUS_IRQ + i;
1277         sysbus_connect_irq(sysbus, i, qdev_get_gpio_in(irqchip, irq));
1278     }
1279 
1280     memory_region_add_subregion(sysmem,
1281                                 memmap[VIRT_PLATFORM_BUS].base,
1282                                 sysbus_mmio_get_region(sysbus, 0));
1283 }
1284 
1285 static void virt_build_smbios(RISCVVirtState *s)
1286 {
1287     MachineClass *mc = MACHINE_GET_CLASS(s);
1288     MachineState *ms = MACHINE(s);
1289     uint8_t *smbios_tables, *smbios_anchor;
1290     size_t smbios_tables_len, smbios_anchor_len;
1291     struct smbios_phys_mem_area mem_array;
1292     const char *product = "QEMU Virtual Machine";
1293 
1294     if (kvm_enabled()) {
1295         product = "KVM Virtual Machine";
1296     }
1297 
1298     smbios_set_defaults("QEMU", product, mc->name);
1299 
1300     if (riscv_is_32bit(&s->soc[0])) {
1301         smbios_set_default_processor_family(0x200);
1302     } else {
1303         smbios_set_default_processor_family(0x201);
1304     }
1305 
1306     /* build the array of physical mem area from base_memmap */
1307     mem_array.address = s->memmap[VIRT_DRAM].base;
1308     mem_array.length = ms->ram_size;
1309 
1310     smbios_get_tables(ms, SMBIOS_ENTRY_POINT_TYPE_64,
1311                       &mem_array, 1,
1312                       &smbios_tables, &smbios_tables_len,
1313                       &smbios_anchor, &smbios_anchor_len,
1314                       &error_fatal);
1315 
1316     if (smbios_anchor) {
1317         fw_cfg_add_file(s->fw_cfg, "etc/smbios/smbios-tables",
1318                         smbios_tables, smbios_tables_len);
1319         fw_cfg_add_file(s->fw_cfg, "etc/smbios/smbios-anchor",
1320                         smbios_anchor, smbios_anchor_len);
1321     }
1322 }
1323 
1324 static void virt_machine_done(Notifier *notifier, void *data)
1325 {
1326     RISCVVirtState *s = container_of(notifier, RISCVVirtState,
1327                                      machine_done);
1328     const MemMapEntry *memmap = virt_memmap;
1329     MachineState *machine = MACHINE(s);
1330     target_ulong start_addr = memmap[VIRT_DRAM].base;
1331     target_ulong firmware_end_addr, kernel_start_addr;
1332     const char *firmware_name = riscv_default_firmware_name(&s->soc[0]);
1333     uint64_t fdt_load_addr;
1334     uint64_t kernel_entry = 0;
1335     BlockBackend *pflash_blk0;
1336 
1337     /*
1338      * An user provided dtb must include everything, including
1339      * dynamic sysbus devices. Our FDT needs to be finalized.
1340      */
1341     if (machine->dtb == NULL) {
1342         finalize_fdt(s);
1343     }
1344 
1345     /*
1346      * Only direct boot kernel is currently supported for KVM VM,
1347      * so the "-bios" parameter is not supported when KVM is enabled.
1348      */
1349     if (kvm_enabled()) {
1350         if (machine->firmware) {
1351             if (strcmp(machine->firmware, "none")) {
1352                 error_report("Machine mode firmware is not supported in "
1353                              "combination with KVM.");
1354                 exit(1);
1355             }
1356         } else {
1357             machine->firmware = g_strdup("none");
1358         }
1359     }
1360 
1361     firmware_end_addr = riscv_find_and_load_firmware(machine, firmware_name,
1362                                                      start_addr, NULL);
1363 
1364     pflash_blk0 = pflash_cfi01_get_blk(s->flash[0]);
1365     if (pflash_blk0) {
1366         if (machine->firmware && !strcmp(machine->firmware, "none") &&
1367             !kvm_enabled()) {
1368             /*
1369              * Pflash was supplied but bios is none and not KVM guest,
1370              * let's overwrite the address we jump to after reset to
1371              * the base of the flash.
1372              */
1373             start_addr = virt_memmap[VIRT_FLASH].base;
1374         } else {
1375             /*
1376              * Pflash was supplied but either KVM guest or bios is not none.
1377              * In this case, base of the flash would contain S-mode payload.
1378              */
1379             riscv_setup_firmware_boot(machine);
1380             kernel_entry = virt_memmap[VIRT_FLASH].base;
1381         }
1382     }
1383 
1384     if (machine->kernel_filename && !kernel_entry) {
1385         kernel_start_addr = riscv_calc_kernel_start_addr(&s->soc[0],
1386                                                          firmware_end_addr);
1387 
1388         kernel_entry = riscv_load_kernel(machine, &s->soc[0],
1389                                          kernel_start_addr, true, NULL);
1390     }
1391 
1392     fdt_load_addr = riscv_compute_fdt_addr(memmap[VIRT_DRAM].base,
1393                                            memmap[VIRT_DRAM].size,
1394                                            machine);
1395     riscv_load_fdt(fdt_load_addr, machine->fdt);
1396 
1397     /* load the reset vector */
1398     riscv_setup_rom_reset_vec(machine, &s->soc[0], start_addr,
1399                               virt_memmap[VIRT_MROM].base,
1400                               virt_memmap[VIRT_MROM].size, kernel_entry,
1401                               fdt_load_addr);
1402 
1403     /*
1404      * Only direct boot kernel is currently supported for KVM VM,
1405      * So here setup kernel start address and fdt address.
1406      * TODO:Support firmware loading and integrate to TCG start
1407      */
1408     if (kvm_enabled()) {
1409         riscv_setup_direct_kernel(kernel_entry, fdt_load_addr);
1410     }
1411 
1412     virt_build_smbios(s);
1413 
1414     if (virt_is_acpi_enabled(s)) {
1415         virt_acpi_setup(s);
1416     }
1417 }
1418 
1419 static void virt_machine_init(MachineState *machine)
1420 {
1421     const MemMapEntry *memmap = virt_memmap;
1422     RISCVVirtState *s = RISCV_VIRT_MACHINE(machine);
1423     MemoryRegion *system_memory = get_system_memory();
1424     MemoryRegion *mask_rom = g_new(MemoryRegion, 1);
1425     DeviceState *mmio_irqchip, *virtio_irqchip, *pcie_irqchip;
1426     int i, base_hartid, hart_count;
1427     int socket_count = riscv_socket_count(machine);
1428 
1429     /* Check socket count limit */
1430     if (VIRT_SOCKETS_MAX < socket_count) {
1431         error_report("number of sockets/nodes should be less than %d",
1432             VIRT_SOCKETS_MAX);
1433         exit(1);
1434     }
1435 
1436     if (!virt_aclint_allowed() && s->have_aclint) {
1437         error_report("'aclint' is only available with TCG acceleration");
1438         exit(1);
1439     }
1440 
1441     /* Initialize sockets */
1442     mmio_irqchip = virtio_irqchip = pcie_irqchip = NULL;
1443     for (i = 0; i < socket_count; i++) {
1444         g_autofree char *soc_name = g_strdup_printf("soc%d", i);
1445 
1446         if (!riscv_socket_check_hartids(machine, i)) {
1447             error_report("discontinuous hartids in socket%d", i);
1448             exit(1);
1449         }
1450 
1451         base_hartid = riscv_socket_first_hartid(machine, i);
1452         if (base_hartid < 0) {
1453             error_report("can't find hartid base for socket%d", i);
1454             exit(1);
1455         }
1456 
1457         hart_count = riscv_socket_hart_count(machine, i);
1458         if (hart_count < 0) {
1459             error_report("can't find hart count for socket%d", i);
1460             exit(1);
1461         }
1462 
1463         object_initialize_child(OBJECT(machine), soc_name, &s->soc[i],
1464                                 TYPE_RISCV_HART_ARRAY);
1465         object_property_set_str(OBJECT(&s->soc[i]), "cpu-type",
1466                                 machine->cpu_type, &error_abort);
1467         object_property_set_int(OBJECT(&s->soc[i]), "hartid-base",
1468                                 base_hartid, &error_abort);
1469         object_property_set_int(OBJECT(&s->soc[i]), "num-harts",
1470                                 hart_count, &error_abort);
1471         sysbus_realize(SYS_BUS_DEVICE(&s->soc[i]), &error_fatal);
1472 
1473         if (virt_aclint_allowed() && s->have_aclint) {
1474             if (s->aia_type == VIRT_AIA_TYPE_APLIC_IMSIC) {
1475                 /* Per-socket ACLINT MTIMER */
1476                 riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1477                             i * RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1478                         RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1479                         base_hartid, hart_count,
1480                         RISCV_ACLINT_DEFAULT_MTIMECMP,
1481                         RISCV_ACLINT_DEFAULT_MTIME,
1482                         RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1483             } else {
1484                 /* Per-socket ACLINT MSWI, MTIMER, and SSWI */
1485                 riscv_aclint_swi_create(memmap[VIRT_CLINT].base +
1486                             i * memmap[VIRT_CLINT].size,
1487                         base_hartid, hart_count, false);
1488                 riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1489                             i * memmap[VIRT_CLINT].size +
1490                             RISCV_ACLINT_SWI_SIZE,
1491                         RISCV_ACLINT_DEFAULT_MTIMER_SIZE,
1492                         base_hartid, hart_count,
1493                         RISCV_ACLINT_DEFAULT_MTIMECMP,
1494                         RISCV_ACLINT_DEFAULT_MTIME,
1495                         RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1496                 riscv_aclint_swi_create(memmap[VIRT_ACLINT_SSWI].base +
1497                             i * memmap[VIRT_ACLINT_SSWI].size,
1498                         base_hartid, hart_count, true);
1499             }
1500         } else if (tcg_enabled()) {
1501             /* Per-socket SiFive CLINT */
1502             riscv_aclint_swi_create(
1503                     memmap[VIRT_CLINT].base + i * memmap[VIRT_CLINT].size,
1504                     base_hartid, hart_count, false);
1505             riscv_aclint_mtimer_create(memmap[VIRT_CLINT].base +
1506                         i * memmap[VIRT_CLINT].size + RISCV_ACLINT_SWI_SIZE,
1507                     RISCV_ACLINT_DEFAULT_MTIMER_SIZE, base_hartid, hart_count,
1508                     RISCV_ACLINT_DEFAULT_MTIMECMP, RISCV_ACLINT_DEFAULT_MTIME,
1509                     RISCV_ACLINT_DEFAULT_TIMEBASE_FREQ, true);
1510         }
1511 
1512         /* Per-socket interrupt controller */
1513         if (s->aia_type == VIRT_AIA_TYPE_NONE) {
1514             s->irqchip[i] = virt_create_plic(memmap, i,
1515                                              base_hartid, hart_count);
1516         } else {
1517             s->irqchip[i] = virt_create_aia(s->aia_type, s->aia_guests,
1518                                             memmap, i, base_hartid,
1519                                             hart_count);
1520         }
1521 
1522         /* Try to use different IRQCHIP instance based device type */
1523         if (i == 0) {
1524             mmio_irqchip = s->irqchip[i];
1525             virtio_irqchip = s->irqchip[i];
1526             pcie_irqchip = s->irqchip[i];
1527         }
1528         if (i == 1) {
1529             virtio_irqchip = s->irqchip[i];
1530             pcie_irqchip = s->irqchip[i];
1531         }
1532         if (i == 2) {
1533             pcie_irqchip = s->irqchip[i];
1534         }
1535     }
1536 
1537     if (kvm_enabled() && virt_use_kvm_aia(s)) {
1538         kvm_riscv_aia_create(machine, IMSIC_MMIO_GROUP_MIN_SHIFT,
1539                              VIRT_IRQCHIP_NUM_SOURCES, VIRT_IRQCHIP_NUM_MSIS,
1540                              memmap[VIRT_APLIC_S].base,
1541                              memmap[VIRT_IMSIC_S].base,
1542                              s->aia_guests);
1543     }
1544 
1545     if (riscv_is_32bit(&s->soc[0])) {
1546 #if HOST_LONG_BITS == 64
1547         /* limit RAM size in a 32-bit system */
1548         if (machine->ram_size > 10 * GiB) {
1549             machine->ram_size = 10 * GiB;
1550             error_report("Limiting RAM size to 10 GiB");
1551         }
1552 #endif
1553         virt_high_pcie_memmap.base = VIRT32_HIGH_PCIE_MMIO_BASE;
1554         virt_high_pcie_memmap.size = VIRT32_HIGH_PCIE_MMIO_SIZE;
1555     } else {
1556         virt_high_pcie_memmap.size = VIRT64_HIGH_PCIE_MMIO_SIZE;
1557         virt_high_pcie_memmap.base = memmap[VIRT_DRAM].base + machine->ram_size;
1558         virt_high_pcie_memmap.base =
1559             ROUND_UP(virt_high_pcie_memmap.base, virt_high_pcie_memmap.size);
1560     }
1561 
1562     s->memmap = virt_memmap;
1563 
1564     /* register system main memory (actual RAM) */
1565     memory_region_add_subregion(system_memory, memmap[VIRT_DRAM].base,
1566         machine->ram);
1567 
1568     /* boot rom */
1569     memory_region_init_rom(mask_rom, NULL, "riscv_virt_board.mrom",
1570                            memmap[VIRT_MROM].size, &error_fatal);
1571     memory_region_add_subregion(system_memory, memmap[VIRT_MROM].base,
1572                                 mask_rom);
1573 
1574     /*
1575      * Init fw_cfg. Must be done before riscv_load_fdt, otherwise the
1576      * device tree cannot be altered and we get FDT_ERR_NOSPACE.
1577      */
1578     s->fw_cfg = create_fw_cfg(machine);
1579     rom_set_fw(s->fw_cfg);
1580 
1581     /* SiFive Test MMIO device */
1582     sifive_test_create(memmap[VIRT_TEST].base);
1583 
1584     /* VirtIO MMIO devices */
1585     for (i = 0; i < VIRTIO_COUNT; i++) {
1586         sysbus_create_simple("virtio-mmio",
1587             memmap[VIRT_VIRTIO].base + i * memmap[VIRT_VIRTIO].size,
1588             qdev_get_gpio_in(virtio_irqchip, VIRTIO_IRQ + i));
1589     }
1590 
1591     gpex_pcie_init(system_memory, pcie_irqchip, s);
1592 
1593     create_platform_bus(s, mmio_irqchip);
1594 
1595     serial_mm_init(system_memory, memmap[VIRT_UART0].base,
1596         0, qdev_get_gpio_in(mmio_irqchip, UART0_IRQ), 399193,
1597         serial_hd(0), DEVICE_LITTLE_ENDIAN);
1598 
1599     sysbus_create_simple("goldfish_rtc", memmap[VIRT_RTC].base,
1600         qdev_get_gpio_in(mmio_irqchip, RTC_IRQ));
1601 
1602     for (i = 0; i < ARRAY_SIZE(s->flash); i++) {
1603         /* Map legacy -drive if=pflash to machine properties */
1604         pflash_cfi01_legacy_drive(s->flash[i],
1605                                   drive_get(IF_PFLASH, 0, i));
1606     }
1607     virt_flash_map(s, system_memory);
1608 
1609     /* load/create device tree */
1610     if (machine->dtb) {
1611         machine->fdt = load_device_tree(machine->dtb, &s->fdt_size);
1612         if (!machine->fdt) {
1613             error_report("load_device_tree() failed");
1614             exit(1);
1615         }
1616     } else {
1617         create_fdt(s, memmap);
1618     }
1619 
1620     s->machine_done.notify = virt_machine_done;
1621     qemu_add_machine_init_done_notifier(&s->machine_done);
1622 }
1623 
1624 static void virt_machine_instance_init(Object *obj)
1625 {
1626     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1627 
1628     virt_flash_create(s);
1629 
1630     s->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
1631     s->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
1632     s->acpi = ON_OFF_AUTO_AUTO;
1633 }
1634 
1635 static char *virt_get_aia_guests(Object *obj, Error **errp)
1636 {
1637     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1638 
1639     return g_strdup_printf("%d", s->aia_guests);
1640 }
1641 
1642 static void virt_set_aia_guests(Object *obj, const char *val, Error **errp)
1643 {
1644     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1645 
1646     s->aia_guests = atoi(val);
1647     if (s->aia_guests < 0 || s->aia_guests > VIRT_IRQCHIP_MAX_GUESTS) {
1648         error_setg(errp, "Invalid number of AIA IMSIC guests");
1649         error_append_hint(errp, "Valid values be between 0 and %d.\n",
1650                           VIRT_IRQCHIP_MAX_GUESTS);
1651     }
1652 }
1653 
1654 static char *virt_get_aia(Object *obj, Error **errp)
1655 {
1656     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1657     const char *val;
1658 
1659     switch (s->aia_type) {
1660     case VIRT_AIA_TYPE_APLIC:
1661         val = "aplic";
1662         break;
1663     case VIRT_AIA_TYPE_APLIC_IMSIC:
1664         val = "aplic-imsic";
1665         break;
1666     default:
1667         val = "none";
1668         break;
1669     };
1670 
1671     return g_strdup(val);
1672 }
1673 
1674 static void virt_set_aia(Object *obj, const char *val, Error **errp)
1675 {
1676     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1677 
1678     if (!strcmp(val, "none")) {
1679         s->aia_type = VIRT_AIA_TYPE_NONE;
1680     } else if (!strcmp(val, "aplic")) {
1681         s->aia_type = VIRT_AIA_TYPE_APLIC;
1682     } else if (!strcmp(val, "aplic-imsic")) {
1683         s->aia_type = VIRT_AIA_TYPE_APLIC_IMSIC;
1684     } else {
1685         error_setg(errp, "Invalid AIA interrupt controller type");
1686         error_append_hint(errp, "Valid values are none, aplic, and "
1687                           "aplic-imsic.\n");
1688     }
1689 }
1690 
1691 static bool virt_get_aclint(Object *obj, Error **errp)
1692 {
1693     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1694 
1695     return s->have_aclint;
1696 }
1697 
1698 static void virt_set_aclint(Object *obj, bool value, Error **errp)
1699 {
1700     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1701 
1702     s->have_aclint = value;
1703 }
1704 
1705 bool virt_is_acpi_enabled(RISCVVirtState *s)
1706 {
1707     return s->acpi != ON_OFF_AUTO_OFF;
1708 }
1709 
1710 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
1711                           void *opaque, Error **errp)
1712 {
1713     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1714     OnOffAuto acpi = s->acpi;
1715 
1716     visit_type_OnOffAuto(v, name, &acpi, errp);
1717 }
1718 
1719 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
1720                           void *opaque, Error **errp)
1721 {
1722     RISCVVirtState *s = RISCV_VIRT_MACHINE(obj);
1723 
1724     visit_type_OnOffAuto(v, name, &s->acpi, errp);
1725 }
1726 
1727 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
1728                                                         DeviceState *dev)
1729 {
1730     MachineClass *mc = MACHINE_GET_CLASS(machine);
1731 
1732     if (device_is_dynamic_sysbus(mc, dev) ||
1733         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1734         return HOTPLUG_HANDLER(machine);
1735     }
1736     return NULL;
1737 }
1738 
1739 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1740                                         DeviceState *dev, Error **errp)
1741 {
1742     RISCVVirtState *s = RISCV_VIRT_MACHINE(hotplug_dev);
1743 
1744     if (s->platform_bus_dev) {
1745         MachineClass *mc = MACHINE_GET_CLASS(s);
1746 
1747         if (device_is_dynamic_sysbus(mc, dev)) {
1748             platform_bus_link_device(PLATFORM_BUS_DEVICE(s->platform_bus_dev),
1749                                      SYS_BUS_DEVICE(dev));
1750         }
1751     }
1752 
1753     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1754         create_fdt_virtio_iommu(s, pci_get_bdf(PCI_DEVICE(dev)));
1755     }
1756 }
1757 
1758 static void virt_machine_class_init(ObjectClass *oc, void *data)
1759 {
1760     MachineClass *mc = MACHINE_CLASS(oc);
1761     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1762 
1763     mc->desc = "RISC-V VirtIO board";
1764     mc->init = virt_machine_init;
1765     mc->max_cpus = VIRT_CPUS_MAX;
1766     mc->default_cpu_type = TYPE_RISCV_CPU_BASE;
1767     mc->block_default_type = IF_VIRTIO;
1768     mc->no_cdrom = 1;
1769     mc->pci_allow_0_address = true;
1770     mc->possible_cpu_arch_ids = riscv_numa_possible_cpu_arch_ids;
1771     mc->cpu_index_to_instance_props = riscv_numa_cpu_index_to_props;
1772     mc->get_default_cpu_node_id = riscv_numa_get_default_cpu_node_id;
1773     mc->numa_mem_supported = true;
1774     /* platform instead of architectural choice */
1775     mc->cpu_cluster_has_numa_boundary = true;
1776     mc->default_ram_id = "riscv_virt_board.ram";
1777     assert(!mc->get_hotplug_handler);
1778     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
1779 
1780     hc->plug = virt_machine_device_plug_cb;
1781 
1782     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
1783 #ifdef CONFIG_TPM
1784     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
1785 #endif
1786 
1787     object_class_property_add_bool(oc, "aclint", virt_get_aclint,
1788                                    virt_set_aclint);
1789     object_class_property_set_description(oc, "aclint",
1790                                           "(TCG only) Set on/off to "
1791                                           "enable/disable emulating "
1792                                           "ACLINT devices");
1793 
1794     object_class_property_add_str(oc, "aia", virt_get_aia,
1795                                   virt_set_aia);
1796     object_class_property_set_description(oc, "aia",
1797                                           "Set type of AIA interrupt "
1798                                           "controller. Valid values are "
1799                                           "none, aplic, and aplic-imsic.");
1800 
1801     object_class_property_add_str(oc, "aia-guests",
1802                                   virt_get_aia_guests,
1803                                   virt_set_aia_guests);
1804     {
1805         g_autofree char *str =
1806             g_strdup_printf("Set number of guest MMIO pages for AIA IMSIC. "
1807                             "Valid value should be between 0 and %d.",
1808                             VIRT_IRQCHIP_MAX_GUESTS);
1809         object_class_property_set_description(oc, "aia-guests", str);
1810     }
1811 
1812     object_class_property_add(oc, "acpi", "OnOffAuto",
1813                               virt_get_acpi, virt_set_acpi,
1814                               NULL, NULL);
1815     object_class_property_set_description(oc, "acpi",
1816                                           "Enable ACPI");
1817 }
1818 
1819 static const TypeInfo virt_machine_typeinfo = {
1820     .name       = MACHINE_TYPE_NAME("virt"),
1821     .parent     = TYPE_MACHINE,
1822     .class_init = virt_machine_class_init,
1823     .instance_init = virt_machine_instance_init,
1824     .instance_size = sizeof(RISCVVirtState),
1825     .interfaces = (InterfaceInfo[]) {
1826          { TYPE_HOTPLUG_HANDLER },
1827          { }
1828     },
1829 };
1830 
1831 static void virt_machine_init_register_types(void)
1832 {
1833     type_register_static(&virt_machine_typeinfo);
1834 }
1835 
1836 type_init(virt_machine_init_register_types)
1837