xref: /openbmc/qemu/hw/arm/virt.c (revision e3a99063)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu-common.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "qapi/error.h"
37 #include "hw/sysbus.h"
38 #include "hw/boards.h"
39 #include "hw/arm/boot.h"
40 #include "hw/arm/primecell.h"
41 #include "hw/arm/virt.h"
42 #include "hw/block/flash.h"
43 #include "hw/vfio/vfio-calxeda-xgmac.h"
44 #include "hw/vfio/vfio-amd-xgbe.h"
45 #include "hw/display/ramfb.h"
46 #include "net/net.h"
47 #include "sysemu/device_tree.h"
48 #include "sysemu/numa.h"
49 #include "sysemu/runstate.h"
50 #include "sysemu/sysemu.h"
51 #include "sysemu/tpm.h"
52 #include "sysemu/kvm.h"
53 #include "hw/loader.h"
54 #include "exec/address-spaces.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/arm/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/irq.h"
67 #include "kvm_arm.h"
68 #include "hw/firmware/smbios.h"
69 #include "qapi/visitor.h"
70 #include "qapi/qapi-visit-common.h"
71 #include "standard-headers/linux/input.h"
72 #include "hw/arm/smmuv3.h"
73 #include "hw/acpi/acpi.h"
74 #include "target/arm/internals.h"
75 #include "hw/mem/pc-dimm.h"
76 #include "hw/mem/nvdimm.h"
77 #include "hw/acpi/generic_event_device.h"
78 #include "hw/virtio/virtio-iommu.h"
79 #include "hw/char/pl011.h"
80 #include "qemu/guest-random.h"
81 
82 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
83     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
84                                                     void *data) \
85     { \
86         MachineClass *mc = MACHINE_CLASS(oc); \
87         virt_machine_##major##_##minor##_options(mc); \
88         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
89         if (latest) { \
90             mc->alias = "virt"; \
91         } \
92     } \
93     static const TypeInfo machvirt_##major##_##minor##_info = { \
94         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
95         .parent = TYPE_VIRT_MACHINE, \
96         .class_init = virt_##major##_##minor##_class_init, \
97     }; \
98     static void machvirt_machine_##major##_##minor##_init(void) \
99     { \
100         type_register_static(&machvirt_##major##_##minor##_info); \
101     } \
102     type_init(machvirt_machine_##major##_##minor##_init);
103 
104 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
105     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
106 #define DEFINE_VIRT_MACHINE(major, minor) \
107     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
108 
109 
110 /* Number of external interrupt lines to configure the GIC with */
111 #define NUM_IRQS 256
112 
113 #define PLATFORM_BUS_NUM_IRQS 64
114 
115 /* Legacy RAM limit in GB (< version 4.0) */
116 #define LEGACY_RAMLIMIT_GB 255
117 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
118 
119 /* Addresses and sizes of our components.
120  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
121  * 128MB..256MB is used for miscellaneous device I/O.
122  * 256MB..1GB is reserved for possible future PCI support (ie where the
123  * PCI memory window will go if we add a PCI host controller).
124  * 1GB and up is RAM (which may happily spill over into the
125  * high memory region beyond 4GB).
126  * This represents a compromise between how much RAM can be given to
127  * a 32 bit VM and leaving space for expansion and in particular for PCI.
128  * Note that devices should generally be placed at multiples of 0x10000,
129  * to accommodate guests using 64K pages.
130  */
131 static const MemMapEntry base_memmap[] = {
132     /* Space up to 0x8000000 is reserved for a boot ROM */
133     [VIRT_FLASH] =              {          0, 0x08000000 },
134     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
135     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
136     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
137     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
138     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
139     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
140     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
141     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
142     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
143     /* This redistributor space allows up to 2*64kB*123 CPUs */
144     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
145     [VIRT_UART] =               { 0x09000000, 0x00001000 },
146     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
147     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
148     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
149     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
150     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
151     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
152     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
153     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
154     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
155     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
156     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
157     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
158     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
159     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
160     /* Actual RAM size depends on initial RAM and device memory settings */
161     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
162 };
163 
164 /*
165  * Highmem IO Regions: This memory map is floating, located after the RAM.
166  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
167  * top of the RAM, so that its base get the same alignment as the size,
168  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
169  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
170  * Note the extended_memmap is sized so that it eventually also includes the
171  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
172  * index of base_memmap).
173  */
174 static MemMapEntry extended_memmap[] = {
175     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
176     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
177     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
178     /* Second PCIe window */
179     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
180 };
181 
182 static const int a15irqmap[] = {
183     [VIRT_UART] = 1,
184     [VIRT_RTC] = 2,
185     [VIRT_PCIE] = 3, /* ... to 6 */
186     [VIRT_GPIO] = 7,
187     [VIRT_SECURE_UART] = 8,
188     [VIRT_ACPI_GED] = 9,
189     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
190     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
191     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
192     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
193 };
194 
195 static const char *valid_cpus[] = {
196     ARM_CPU_TYPE_NAME("cortex-a7"),
197     ARM_CPU_TYPE_NAME("cortex-a15"),
198     ARM_CPU_TYPE_NAME("cortex-a53"),
199     ARM_CPU_TYPE_NAME("cortex-a57"),
200     ARM_CPU_TYPE_NAME("cortex-a72"),
201     ARM_CPU_TYPE_NAME("host"),
202     ARM_CPU_TYPE_NAME("max"),
203 };
204 
205 static bool cpu_type_valid(const char *cpu)
206 {
207     int i;
208 
209     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
210         if (strcmp(cpu, valid_cpus[i]) == 0) {
211             return true;
212         }
213     }
214     return false;
215 }
216 
217 static void create_kaslr_seed(VirtMachineState *vms, const char *node)
218 {
219     Error *err = NULL;
220     uint64_t seed;
221 
222     if (qemu_guest_getrandom(&seed, sizeof(seed), &err)) {
223         error_free(err);
224         return;
225     }
226     qemu_fdt_setprop_u64(vms->fdt, node, "kaslr-seed", seed);
227 }
228 
229 static void create_fdt(VirtMachineState *vms)
230 {
231     MachineState *ms = MACHINE(vms);
232     int nb_numa_nodes = ms->numa_state->num_nodes;
233     void *fdt = create_device_tree(&vms->fdt_size);
234 
235     if (!fdt) {
236         error_report("create_device_tree() failed");
237         exit(1);
238     }
239 
240     vms->fdt = fdt;
241 
242     /* Header */
243     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
244     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
245     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
246 
247     /* /chosen must exist for load_dtb to fill in necessary properties later */
248     qemu_fdt_add_subnode(fdt, "/chosen");
249     create_kaslr_seed(vms, "/chosen");
250 
251     if (vms->secure) {
252         qemu_fdt_add_subnode(fdt, "/secure-chosen");
253         create_kaslr_seed(vms, "/secure-chosen");
254     }
255 
256     /* Clock node, for the benefit of the UART. The kernel device tree
257      * binding documentation claims the PL011 node clock properties are
258      * optional but in practice if you omit them the kernel refuses to
259      * probe for the device.
260      */
261     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
262     qemu_fdt_add_subnode(fdt, "/apb-pclk");
263     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
264     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
265     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
266     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
267                                 "clk24mhz");
268     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
269 
270     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
271         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
272         uint32_t *matrix = g_malloc0(size);
273         int idx, i, j;
274 
275         for (i = 0; i < nb_numa_nodes; i++) {
276             for (j = 0; j < nb_numa_nodes; j++) {
277                 idx = (i * nb_numa_nodes + j) * 3;
278                 matrix[idx + 0] = cpu_to_be32(i);
279                 matrix[idx + 1] = cpu_to_be32(j);
280                 matrix[idx + 2] =
281                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
282             }
283         }
284 
285         qemu_fdt_add_subnode(fdt, "/distance-map");
286         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
287                                 "numa-distance-map-v1");
288         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
289                          matrix, size);
290         g_free(matrix);
291     }
292 }
293 
294 static void fdt_add_timer_nodes(const VirtMachineState *vms)
295 {
296     /* On real hardware these interrupts are level-triggered.
297      * On KVM they were edge-triggered before host kernel version 4.4,
298      * and level-triggered afterwards.
299      * On emulated QEMU they are level-triggered.
300      *
301      * Getting the DTB info about them wrong is awkward for some
302      * guest kernels:
303      *  pre-4.8 ignore the DT and leave the interrupt configured
304      *   with whatever the GIC reset value (or the bootloader) left it at
305      *  4.8 before rc6 honour the incorrect data by programming it back
306      *   into the GIC, causing problems
307      *  4.8rc6 and later ignore the DT and always write "level triggered"
308      *   into the GIC
309      *
310      * For backwards-compatibility, virt-2.8 and earlier will continue
311      * to say these are edge-triggered, but later machines will report
312      * the correct information.
313      */
314     ARMCPU *armcpu;
315     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
316     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
317 
318     if (vmc->claim_edge_triggered_timers) {
319         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
320     }
321 
322     if (vms->gic_version == VIRT_GIC_VERSION_2) {
323         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
324                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
325                              (1 << vms->smp_cpus) - 1);
326     }
327 
328     qemu_fdt_add_subnode(vms->fdt, "/timer");
329 
330     armcpu = ARM_CPU(qemu_get_cpu(0));
331     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
332         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
333         qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
334                          compat, sizeof(compat));
335     } else {
336         qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
337                                 "arm,armv7-timer");
338     }
339     qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
340     qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
341                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
342                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
343                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
344                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
345 }
346 
347 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
348 {
349     int cpu;
350     int addr_cells = 1;
351     const MachineState *ms = MACHINE(vms);
352 
353     /*
354      * From Documentation/devicetree/bindings/arm/cpus.txt
355      *  On ARM v8 64-bit systems value should be set to 2,
356      *  that corresponds to the MPIDR_EL1 register size.
357      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
358      *  in the system, #address-cells can be set to 1, since
359      *  MPIDR_EL1[63:32] bits are not used for CPUs
360      *  identification.
361      *
362      *  Here we actually don't know whether our system is 32- or 64-bit one.
363      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
364      *  at least one of them has Aff3 populated, we set #address-cells to 2.
365      */
366     for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
367         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
368 
369         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
370             addr_cells = 2;
371             break;
372         }
373     }
374 
375     qemu_fdt_add_subnode(vms->fdt, "/cpus");
376     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
377     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
378 
379     for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
380         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
381         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
382         CPUState *cs = CPU(armcpu);
383 
384         qemu_fdt_add_subnode(vms->fdt, nodename);
385         qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
386         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
387                                     armcpu->dtb_compatible);
388 
389         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
390             && vms->smp_cpus > 1) {
391             qemu_fdt_setprop_string(vms->fdt, nodename,
392                                         "enable-method", "psci");
393         }
394 
395         if (addr_cells == 2) {
396             qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
397                                  armcpu->mp_affinity);
398         } else {
399             qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
400                                   armcpu->mp_affinity);
401         }
402 
403         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
404             qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
405                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
406         }
407 
408         g_free(nodename);
409     }
410 }
411 
412 static void fdt_add_its_gic_node(VirtMachineState *vms)
413 {
414     char *nodename;
415 
416     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
417     nodename = g_strdup_printf("/intc/its@%" PRIx64,
418                                vms->memmap[VIRT_GIC_ITS].base);
419     qemu_fdt_add_subnode(vms->fdt, nodename);
420     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
421                             "arm,gic-v3-its");
422     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
423     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
424                                  2, vms->memmap[VIRT_GIC_ITS].base,
425                                  2, vms->memmap[VIRT_GIC_ITS].size);
426     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
427     g_free(nodename);
428 }
429 
430 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
431 {
432     char *nodename;
433 
434     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
435                                vms->memmap[VIRT_GIC_V2M].base);
436     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
437     qemu_fdt_add_subnode(vms->fdt, nodename);
438     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
439                             "arm,gic-v2m-frame");
440     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
441     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
442                                  2, vms->memmap[VIRT_GIC_V2M].base,
443                                  2, vms->memmap[VIRT_GIC_V2M].size);
444     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
445     g_free(nodename);
446 }
447 
448 static void fdt_add_gic_node(VirtMachineState *vms)
449 {
450     char *nodename;
451 
452     vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
453     qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
454 
455     nodename = g_strdup_printf("/intc@%" PRIx64,
456                                vms->memmap[VIRT_GIC_DIST].base);
457     qemu_fdt_add_subnode(vms->fdt, nodename);
458     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 3);
459     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-controller", NULL, 0);
460     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 0x2);
461     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 0x2);
462     qemu_fdt_setprop(vms->fdt, nodename, "ranges", NULL, 0);
463     if (vms->gic_version == VIRT_GIC_VERSION_3) {
464         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
465 
466         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
467                                 "arm,gic-v3");
468 
469         qemu_fdt_setprop_cell(vms->fdt, nodename,
470                               "#redistributor-regions", nb_redist_regions);
471 
472         if (nb_redist_regions == 1) {
473             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
474                                          2, vms->memmap[VIRT_GIC_DIST].base,
475                                          2, vms->memmap[VIRT_GIC_DIST].size,
476                                          2, vms->memmap[VIRT_GIC_REDIST].base,
477                                          2, vms->memmap[VIRT_GIC_REDIST].size);
478         } else {
479             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
480                                  2, vms->memmap[VIRT_GIC_DIST].base,
481                                  2, vms->memmap[VIRT_GIC_DIST].size,
482                                  2, vms->memmap[VIRT_GIC_REDIST].base,
483                                  2, vms->memmap[VIRT_GIC_REDIST].size,
484                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
485                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
486         }
487 
488         if (vms->virt) {
489             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
490                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
491                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
492         }
493     } else {
494         /* 'cortex-a15-gic' means 'GIC v2' */
495         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
496                                 "arm,cortex-a15-gic");
497         if (!vms->virt) {
498             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
499                                          2, vms->memmap[VIRT_GIC_DIST].base,
500                                          2, vms->memmap[VIRT_GIC_DIST].size,
501                                          2, vms->memmap[VIRT_GIC_CPU].base,
502                                          2, vms->memmap[VIRT_GIC_CPU].size);
503         } else {
504             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
505                                          2, vms->memmap[VIRT_GIC_DIST].base,
506                                          2, vms->memmap[VIRT_GIC_DIST].size,
507                                          2, vms->memmap[VIRT_GIC_CPU].base,
508                                          2, vms->memmap[VIRT_GIC_CPU].size,
509                                          2, vms->memmap[VIRT_GIC_HYP].base,
510                                          2, vms->memmap[VIRT_GIC_HYP].size,
511                                          2, vms->memmap[VIRT_GIC_VCPU].base,
512                                          2, vms->memmap[VIRT_GIC_VCPU].size);
513             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
514                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
515                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
516         }
517     }
518 
519     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->gic_phandle);
520     g_free(nodename);
521 }
522 
523 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
524 {
525     CPUState *cpu;
526     ARMCPU *armcpu;
527     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
528 
529     CPU_FOREACH(cpu) {
530         armcpu = ARM_CPU(cpu);
531         if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
532             return;
533         }
534         if (kvm_enabled()) {
535             if (kvm_irqchip_in_kernel()) {
536                 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
537             }
538             kvm_arm_pmu_init(cpu);
539         }
540     }
541 
542     if (vms->gic_version == VIRT_GIC_VERSION_2) {
543         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
544                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
545                              (1 << vms->smp_cpus) - 1);
546     }
547 
548     armcpu = ARM_CPU(qemu_get_cpu(0));
549     qemu_fdt_add_subnode(vms->fdt, "/pmu");
550     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
551         const char compat[] = "arm,armv8-pmuv3";
552         qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
553                          compat, sizeof(compat));
554         qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
555                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
556     }
557 }
558 
559 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
560 {
561     DeviceState *dev;
562     MachineState *ms = MACHINE(vms);
563     int irq = vms->irqmap[VIRT_ACPI_GED];
564     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
565 
566     if (ms->ram_slots) {
567         event |= ACPI_GED_MEM_HOTPLUG_EVT;
568     }
569 
570     dev = qdev_create(NULL, TYPE_ACPI_GED);
571     qdev_prop_set_uint32(dev, "ged-event", event);
572 
573     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
574     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
575     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
576 
577     qdev_init_nofail(dev);
578 
579     return dev;
580 }
581 
582 static void create_its(VirtMachineState *vms)
583 {
584     const char *itsclass = its_class_name();
585     DeviceState *dev;
586 
587     if (!itsclass) {
588         /* Do nothing if not supported */
589         return;
590     }
591 
592     dev = qdev_create(NULL, itsclass);
593 
594     object_property_set_link(OBJECT(dev), OBJECT(vms->gic), "parent-gicv3",
595                              &error_abort);
596     qdev_init_nofail(dev);
597     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
598 
599     fdt_add_its_gic_node(vms);
600 }
601 
602 static void create_v2m(VirtMachineState *vms)
603 {
604     int i;
605     int irq = vms->irqmap[VIRT_GIC_V2M];
606     DeviceState *dev;
607 
608     dev = qdev_create(NULL, "arm-gicv2m");
609     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
610     qdev_prop_set_uint32(dev, "base-spi", irq);
611     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
612     qdev_init_nofail(dev);
613 
614     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
615         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
616                            qdev_get_gpio_in(vms->gic, irq + i));
617     }
618 
619     fdt_add_v2m_gic_node(vms);
620 }
621 
622 static void create_gic(VirtMachineState *vms)
623 {
624     MachineState *ms = MACHINE(vms);
625     /* We create a standalone GIC */
626     SysBusDevice *gicbusdev;
627     const char *gictype;
628     int type = vms->gic_version, i;
629     unsigned int smp_cpus = ms->smp.cpus;
630     uint32_t nb_redist_regions = 0;
631 
632     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
633 
634     vms->gic = qdev_create(NULL, gictype);
635     qdev_prop_set_uint32(vms->gic, "revision", type);
636     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
637     /* Note that the num-irq property counts both internal and external
638      * interrupts; there are always 32 of the former (mandated by GIC spec).
639      */
640     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
641     if (!kvm_irqchip_in_kernel()) {
642         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
643     }
644 
645     if (type == 3) {
646         uint32_t redist0_capacity =
647                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
648         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
649 
650         nb_redist_regions = virt_gicv3_redist_region_count(vms);
651 
652         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
653                              nb_redist_regions);
654         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
655 
656         if (nb_redist_regions == 2) {
657             uint32_t redist1_capacity =
658                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
659 
660             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
661                 MIN(smp_cpus - redist0_count, redist1_capacity));
662         }
663     } else {
664         if (!kvm_irqchip_in_kernel()) {
665             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
666                               vms->virt);
667         }
668     }
669     qdev_init_nofail(vms->gic);
670     gicbusdev = SYS_BUS_DEVICE(vms->gic);
671     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
672     if (type == 3) {
673         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
674         if (nb_redist_regions == 2) {
675             sysbus_mmio_map(gicbusdev, 2,
676                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
677         }
678     } else {
679         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
680         if (vms->virt) {
681             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
682             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
683         }
684     }
685 
686     /* Wire the outputs from each CPU's generic timer and the GICv3
687      * maintenance interrupt signal to the appropriate GIC PPI inputs,
688      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
689      */
690     for (i = 0; i < smp_cpus; i++) {
691         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
692         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
693         int irq;
694         /* Mapping from the output timer irq lines from the CPU to the
695          * GIC PPI inputs we use for the virt board.
696          */
697         const int timer_irq[] = {
698             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
699             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
700             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
701             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
702         };
703 
704         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
705             qdev_connect_gpio_out(cpudev, irq,
706                                   qdev_get_gpio_in(vms->gic,
707                                                    ppibase + timer_irq[irq]));
708         }
709 
710         if (type == 3) {
711             qemu_irq irq = qdev_get_gpio_in(vms->gic,
712                                             ppibase + ARCH_GIC_MAINT_IRQ);
713             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
714                                         0, irq);
715         } else if (vms->virt) {
716             qemu_irq irq = qdev_get_gpio_in(vms->gic,
717                                             ppibase + ARCH_GIC_MAINT_IRQ);
718             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
719         }
720 
721         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
722                                     qdev_get_gpio_in(vms->gic, ppibase
723                                                      + VIRTUAL_PMU_IRQ));
724 
725         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
726         sysbus_connect_irq(gicbusdev, i + smp_cpus,
727                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
728         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
729                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
730         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
731                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
732     }
733 
734     fdt_add_gic_node(vms);
735 
736     if (type == 3 && vms->its) {
737         create_its(vms);
738     } else if (type == 2) {
739         create_v2m(vms);
740     }
741 }
742 
743 static void create_uart(const VirtMachineState *vms, int uart,
744                         MemoryRegion *mem, Chardev *chr)
745 {
746     char *nodename;
747     hwaddr base = vms->memmap[uart].base;
748     hwaddr size = vms->memmap[uart].size;
749     int irq = vms->irqmap[uart];
750     const char compat[] = "arm,pl011\0arm,primecell";
751     const char clocknames[] = "uartclk\0apb_pclk";
752     DeviceState *dev = qdev_create(NULL, TYPE_PL011);
753     SysBusDevice *s = SYS_BUS_DEVICE(dev);
754 
755     qdev_prop_set_chr(dev, "chardev", chr);
756     qdev_init_nofail(dev);
757     memory_region_add_subregion(mem, base,
758                                 sysbus_mmio_get_region(s, 0));
759     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
760 
761     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
762     qemu_fdt_add_subnode(vms->fdt, nodename);
763     /* Note that we can't use setprop_string because of the embedded NUL */
764     qemu_fdt_setprop(vms->fdt, nodename, "compatible",
765                          compat, sizeof(compat));
766     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
767                                      2, base, 2, size);
768     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
769                                GIC_FDT_IRQ_TYPE_SPI, irq,
770                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
771     qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
772                                vms->clock_phandle, vms->clock_phandle);
773     qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
774                          clocknames, sizeof(clocknames));
775 
776     if (uart == VIRT_UART) {
777         qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
778     } else {
779         /* Mark as not usable by the normal world */
780         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
781         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
782 
783         qemu_fdt_setprop_string(vms->fdt, "/secure-chosen", "stdout-path",
784                                 nodename);
785     }
786 
787     g_free(nodename);
788 }
789 
790 static void create_rtc(const VirtMachineState *vms)
791 {
792     char *nodename;
793     hwaddr base = vms->memmap[VIRT_RTC].base;
794     hwaddr size = vms->memmap[VIRT_RTC].size;
795     int irq = vms->irqmap[VIRT_RTC];
796     const char compat[] = "arm,pl031\0arm,primecell";
797 
798     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
799 
800     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
801     qemu_fdt_add_subnode(vms->fdt, nodename);
802     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
803     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
804                                  2, base, 2, size);
805     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
806                            GIC_FDT_IRQ_TYPE_SPI, irq,
807                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
808     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
809     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
810     g_free(nodename);
811 }
812 
813 static DeviceState *gpio_key_dev;
814 static void virt_powerdown_req(Notifier *n, void *opaque)
815 {
816     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
817 
818     if (s->acpi_dev) {
819         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
820     } else {
821         /* use gpio Pin 3 for power button event */
822         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
823     }
824 }
825 
826 static void create_gpio(const VirtMachineState *vms)
827 {
828     char *nodename;
829     DeviceState *pl061_dev;
830     hwaddr base = vms->memmap[VIRT_GPIO].base;
831     hwaddr size = vms->memmap[VIRT_GPIO].size;
832     int irq = vms->irqmap[VIRT_GPIO];
833     const char compat[] = "arm,pl061\0arm,primecell";
834 
835     pl061_dev = sysbus_create_simple("pl061", base,
836                                      qdev_get_gpio_in(vms->gic, irq));
837 
838     uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
839     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
840     qemu_fdt_add_subnode(vms->fdt, nodename);
841     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
842                                  2, base, 2, size);
843     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
844     qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
845     qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
846     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
847                            GIC_FDT_IRQ_TYPE_SPI, irq,
848                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
849     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
850     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
851     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
852 
853     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
854                                         qdev_get_gpio_in(pl061_dev, 3));
855     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
856     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
857     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
858     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
859 
860     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
861     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
862                             "label", "GPIO Key Poweroff");
863     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
864                           KEY_POWER);
865     qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
866                            "gpios", phandle, 3, 0);
867     g_free(nodename);
868 }
869 
870 static void create_virtio_devices(const VirtMachineState *vms)
871 {
872     int i;
873     hwaddr size = vms->memmap[VIRT_MMIO].size;
874 
875     /* We create the transports in forwards order. Since qbus_realize()
876      * prepends (not appends) new child buses, the incrementing loop below will
877      * create a list of virtio-mmio buses with decreasing base addresses.
878      *
879      * When a -device option is processed from the command line,
880      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
881      * order. The upshot is that -device options in increasing command line
882      * order are mapped to virtio-mmio buses with decreasing base addresses.
883      *
884      * When this code was originally written, that arrangement ensured that the
885      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
886      * the first -device on the command line. (The end-to-end order is a
887      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
888      * guest kernel's name-to-address assignment strategy.)
889      *
890      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
891      * the message, if not necessarily the code, of commit 70161ff336.
892      * Therefore the loop now establishes the inverse of the original intent.
893      *
894      * Unfortunately, we can't counteract the kernel change by reversing the
895      * loop; it would break existing command lines.
896      *
897      * In any case, the kernel makes no guarantee about the stability of
898      * enumeration order of virtio devices (as demonstrated by it changing
899      * between kernel versions). For reliable and stable identification
900      * of disks users must use UUIDs or similar mechanisms.
901      */
902     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
903         int irq = vms->irqmap[VIRT_MMIO] + i;
904         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
905 
906         sysbus_create_simple("virtio-mmio", base,
907                              qdev_get_gpio_in(vms->gic, irq));
908     }
909 
910     /* We add dtb nodes in reverse order so that they appear in the finished
911      * device tree lowest address first.
912      *
913      * Note that this mapping is independent of the loop above. The previous
914      * loop influences virtio device to virtio transport assignment, whereas
915      * this loop controls how virtio transports are laid out in the dtb.
916      */
917     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
918         char *nodename;
919         int irq = vms->irqmap[VIRT_MMIO] + i;
920         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
921 
922         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
923         qemu_fdt_add_subnode(vms->fdt, nodename);
924         qemu_fdt_setprop_string(vms->fdt, nodename,
925                                 "compatible", "virtio,mmio");
926         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
927                                      2, base, 2, size);
928         qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
929                                GIC_FDT_IRQ_TYPE_SPI, irq,
930                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
931         qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
932         g_free(nodename);
933     }
934 }
935 
936 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
937 
938 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
939                                         const char *name,
940                                         const char *alias_prop_name)
941 {
942     /*
943      * Create a single flash device.  We use the same parameters as
944      * the flash devices on the Versatile Express board.
945      */
946     DeviceState *dev = qdev_create(NULL, TYPE_PFLASH_CFI01);
947 
948     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
949     qdev_prop_set_uint8(dev, "width", 4);
950     qdev_prop_set_uint8(dev, "device-width", 2);
951     qdev_prop_set_bit(dev, "big-endian", false);
952     qdev_prop_set_uint16(dev, "id0", 0x89);
953     qdev_prop_set_uint16(dev, "id1", 0x18);
954     qdev_prop_set_uint16(dev, "id2", 0x00);
955     qdev_prop_set_uint16(dev, "id3", 0x00);
956     qdev_prop_set_string(dev, "name", name);
957     object_property_add_child(OBJECT(vms), name, OBJECT(dev),
958                               &error_abort);
959     object_property_add_alias(OBJECT(vms), alias_prop_name,
960                               OBJECT(dev), "drive", &error_abort);
961     return PFLASH_CFI01(dev);
962 }
963 
964 static void virt_flash_create(VirtMachineState *vms)
965 {
966     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
967     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
968 }
969 
970 static void virt_flash_map1(PFlashCFI01 *flash,
971                             hwaddr base, hwaddr size,
972                             MemoryRegion *sysmem)
973 {
974     DeviceState *dev = DEVICE(flash);
975 
976     assert(size % VIRT_FLASH_SECTOR_SIZE == 0);
977     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
978     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
979     qdev_init_nofail(dev);
980 
981     memory_region_add_subregion(sysmem, base,
982                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
983                                                        0));
984 }
985 
986 static void virt_flash_map(VirtMachineState *vms,
987                            MemoryRegion *sysmem,
988                            MemoryRegion *secure_sysmem)
989 {
990     /*
991      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
992      * sysmem is the system memory space. secure_sysmem is the secure view
993      * of the system, and the first flash device should be made visible only
994      * there. The second flash device is visible to both secure and nonsecure.
995      * If sysmem == secure_sysmem this means there is no separate Secure
996      * address space and both flash devices are generally visible.
997      */
998     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
999     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1000 
1001     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1002                     secure_sysmem);
1003     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1004                     sysmem);
1005 }
1006 
1007 static void virt_flash_fdt(VirtMachineState *vms,
1008                            MemoryRegion *sysmem,
1009                            MemoryRegion *secure_sysmem)
1010 {
1011     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1012     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1013     char *nodename;
1014 
1015     if (sysmem == secure_sysmem) {
1016         /* Report both flash devices as a single node in the DT */
1017         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1018         qemu_fdt_add_subnode(vms->fdt, nodename);
1019         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1020         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1021                                      2, flashbase, 2, flashsize,
1022                                      2, flashbase + flashsize, 2, flashsize);
1023         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1024         g_free(nodename);
1025     } else {
1026         /*
1027          * Report the devices as separate nodes so we can mark one as
1028          * only visible to the secure world.
1029          */
1030         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1031         qemu_fdt_add_subnode(vms->fdt, nodename);
1032         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1033         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1034                                      2, flashbase, 2, flashsize);
1035         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1036         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1037         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1038         g_free(nodename);
1039 
1040         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1041         qemu_fdt_add_subnode(vms->fdt, nodename);
1042         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1043         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1044                                      2, flashbase + flashsize, 2, flashsize);
1045         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1046         g_free(nodename);
1047     }
1048 }
1049 
1050 static bool virt_firmware_init(VirtMachineState *vms,
1051                                MemoryRegion *sysmem,
1052                                MemoryRegion *secure_sysmem)
1053 {
1054     int i;
1055     BlockBackend *pflash_blk0;
1056 
1057     /* Map legacy -drive if=pflash to machine properties */
1058     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1059         pflash_cfi01_legacy_drive(vms->flash[i],
1060                                   drive_get(IF_PFLASH, 0, i));
1061     }
1062 
1063     virt_flash_map(vms, sysmem, secure_sysmem);
1064 
1065     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1066 
1067     if (bios_name) {
1068         char *fname;
1069         MemoryRegion *mr;
1070         int image_size;
1071 
1072         if (pflash_blk0) {
1073             error_report("The contents of the first flash device may be "
1074                          "specified with -bios or with -drive if=pflash... "
1075                          "but you cannot use both options at once");
1076             exit(1);
1077         }
1078 
1079         /* Fall back to -bios */
1080 
1081         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1082         if (!fname) {
1083             error_report("Could not find ROM image '%s'", bios_name);
1084             exit(1);
1085         }
1086         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1087         image_size = load_image_mr(fname, mr);
1088         g_free(fname);
1089         if (image_size < 0) {
1090             error_report("Could not load ROM image '%s'", bios_name);
1091             exit(1);
1092         }
1093     }
1094 
1095     return pflash_blk0 || bios_name;
1096 }
1097 
1098 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1099 {
1100     MachineState *ms = MACHINE(vms);
1101     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1102     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1103     FWCfgState *fw_cfg;
1104     char *nodename;
1105 
1106     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1107     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1108 
1109     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1110     qemu_fdt_add_subnode(vms->fdt, nodename);
1111     qemu_fdt_setprop_string(vms->fdt, nodename,
1112                             "compatible", "qemu,fw-cfg-mmio");
1113     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1114                                  2, base, 2, size);
1115     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1116     g_free(nodename);
1117     return fw_cfg;
1118 }
1119 
1120 static void create_pcie_irq_map(const VirtMachineState *vms,
1121                                 uint32_t gic_phandle,
1122                                 int first_irq, const char *nodename)
1123 {
1124     int devfn, pin;
1125     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1126     uint32_t *irq_map = full_irq_map;
1127 
1128     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1129         for (pin = 0; pin < 4; pin++) {
1130             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1131             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1132             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1133             int i;
1134 
1135             uint32_t map[] = {
1136                 devfn << 8, 0, 0,                           /* devfn */
1137                 pin + 1,                                    /* PCI pin */
1138                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1139 
1140             /* Convert map to big endian */
1141             for (i = 0; i < 10; i++) {
1142                 irq_map[i] = cpu_to_be32(map[i]);
1143             }
1144             irq_map += 10;
1145         }
1146     }
1147 
1148     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
1149                      full_irq_map, sizeof(full_irq_map));
1150 
1151     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
1152                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
1153                            0x7           /* PCI irq */);
1154 }
1155 
1156 static void create_smmu(const VirtMachineState *vms,
1157                         PCIBus *bus)
1158 {
1159     char *node;
1160     const char compat[] = "arm,smmu-v3";
1161     int irq =  vms->irqmap[VIRT_SMMU];
1162     int i;
1163     hwaddr base = vms->memmap[VIRT_SMMU].base;
1164     hwaddr size = vms->memmap[VIRT_SMMU].size;
1165     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1166     DeviceState *dev;
1167 
1168     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1169         return;
1170     }
1171 
1172     dev = qdev_create(NULL, "arm-smmuv3");
1173 
1174     object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
1175                              &error_abort);
1176     qdev_init_nofail(dev);
1177     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1178     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1179         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1180                            qdev_get_gpio_in(vms->gic, irq + i));
1181     }
1182 
1183     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1184     qemu_fdt_add_subnode(vms->fdt, node);
1185     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1186     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
1187 
1188     qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
1189             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1190             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1191             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1192             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1193 
1194     qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
1195                      sizeof(irq_names));
1196 
1197     qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
1198     qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
1199     qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
1200 
1201     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1202 
1203     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1204     g_free(node);
1205 }
1206 
1207 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1208 {
1209     const char compat[] = "virtio,pci-iommu";
1210     uint16_t bdf = vms->virtio_iommu_bdf;
1211     char *node;
1212 
1213     vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1214 
1215     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1216     qemu_fdt_add_subnode(vms->fdt, node);
1217     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1218     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg",
1219                                  1, bdf << 8, 1, 0, 1, 0,
1220                                  1, 0, 1, 0);
1221 
1222     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1223     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1224     g_free(node);
1225 
1226     qemu_fdt_setprop_cells(vms->fdt, vms->pciehb_nodename, "iommu-map",
1227                            0x0, vms->iommu_phandle, 0x0, bdf,
1228                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1229 }
1230 
1231 static void create_pcie(VirtMachineState *vms)
1232 {
1233     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1234     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1235     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1236     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1237     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1238     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1239     hwaddr base_ecam, size_ecam;
1240     hwaddr base = base_mmio;
1241     int nr_pcie_buses;
1242     int irq = vms->irqmap[VIRT_PCIE];
1243     MemoryRegion *mmio_alias;
1244     MemoryRegion *mmio_reg;
1245     MemoryRegion *ecam_alias;
1246     MemoryRegion *ecam_reg;
1247     DeviceState *dev;
1248     char *nodename;
1249     int i, ecam_id;
1250     PCIHostState *pci;
1251 
1252     dev = qdev_create(NULL, TYPE_GPEX_HOST);
1253     qdev_init_nofail(dev);
1254 
1255     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1256     base_ecam = vms->memmap[ecam_id].base;
1257     size_ecam = vms->memmap[ecam_id].size;
1258     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1259     /* Map only the first size_ecam bytes of ECAM space */
1260     ecam_alias = g_new0(MemoryRegion, 1);
1261     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1262     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1263                              ecam_reg, 0, size_ecam);
1264     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1265 
1266     /* Map the MMIO window into system address space so as to expose
1267      * the section of PCI MMIO space which starts at the same base address
1268      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1269      * the window).
1270      */
1271     mmio_alias = g_new0(MemoryRegion, 1);
1272     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1273     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1274                              mmio_reg, base_mmio, size_mmio);
1275     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1276 
1277     if (vms->highmem) {
1278         /* Map high MMIO space */
1279         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1280 
1281         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1282                                  mmio_reg, base_mmio_high, size_mmio_high);
1283         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1284                                     high_mmio_alias);
1285     }
1286 
1287     /* Map IO port space */
1288     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1289 
1290     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1291         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1292                            qdev_get_gpio_in(vms->gic, irq + i));
1293         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1294     }
1295 
1296     pci = PCI_HOST_BRIDGE(dev);
1297     if (pci->bus) {
1298         for (i = 0; i < nb_nics; i++) {
1299             NICInfo *nd = &nd_table[i];
1300 
1301             if (!nd->model) {
1302                 nd->model = g_strdup("virtio");
1303             }
1304 
1305             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1306         }
1307     }
1308 
1309     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1310     qemu_fdt_add_subnode(vms->fdt, nodename);
1311     qemu_fdt_setprop_string(vms->fdt, nodename,
1312                             "compatible", "pci-host-ecam-generic");
1313     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1314     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1315     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1316     qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1317     qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1318                            nr_pcie_buses - 1);
1319     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1320 
1321     if (vms->msi_phandle) {
1322         qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1323                                vms->msi_phandle);
1324     }
1325 
1326     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1327                                  2, base_ecam, 2, size_ecam);
1328 
1329     if (vms->highmem) {
1330         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1331                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1332                                      2, base_pio, 2, size_pio,
1333                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1334                                      2, base_mmio, 2, size_mmio,
1335                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1336                                      2, base_mmio_high,
1337                                      2, base_mmio_high, 2, size_mmio_high);
1338     } else {
1339         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1340                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1341                                      2, base_pio, 2, size_pio,
1342                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1343                                      2, base_mmio, 2, size_mmio);
1344     }
1345 
1346     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1347     create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1348 
1349     if (vms->iommu) {
1350         vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1351 
1352         switch (vms->iommu) {
1353         case VIRT_IOMMU_SMMUV3:
1354             create_smmu(vms, pci->bus);
1355             qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1356                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1357             break;
1358         default:
1359             g_assert_not_reached();
1360         }
1361     }
1362 }
1363 
1364 static void create_platform_bus(VirtMachineState *vms)
1365 {
1366     DeviceState *dev;
1367     SysBusDevice *s;
1368     int i;
1369     MemoryRegion *sysmem = get_system_memory();
1370 
1371     dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1372     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1373     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1374     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1375     qdev_init_nofail(dev);
1376     vms->platform_bus_dev = dev;
1377 
1378     s = SYS_BUS_DEVICE(dev);
1379     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1380         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1381         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1382     }
1383 
1384     memory_region_add_subregion(sysmem,
1385                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1386                                 sysbus_mmio_get_region(s, 0));
1387 }
1388 
1389 static void create_secure_ram(VirtMachineState *vms,
1390                               MemoryRegion *secure_sysmem)
1391 {
1392     MemoryRegion *secram = g_new(MemoryRegion, 1);
1393     char *nodename;
1394     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1395     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1396 
1397     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1398                            &error_fatal);
1399     memory_region_add_subregion(secure_sysmem, base, secram);
1400 
1401     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1402     qemu_fdt_add_subnode(vms->fdt, nodename);
1403     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1404     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1405     qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1406     qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1407 
1408     g_free(nodename);
1409 }
1410 
1411 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1412 {
1413     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1414                                                  bootinfo);
1415 
1416     *fdt_size = board->fdt_size;
1417     return board->fdt;
1418 }
1419 
1420 static void virt_build_smbios(VirtMachineState *vms)
1421 {
1422     MachineClass *mc = MACHINE_GET_CLASS(vms);
1423     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1424     uint8_t *smbios_tables, *smbios_anchor;
1425     size_t smbios_tables_len, smbios_anchor_len;
1426     const char *product = "QEMU Virtual Machine";
1427 
1428     if (kvm_enabled()) {
1429         product = "KVM Virtual Machine";
1430     }
1431 
1432     smbios_set_defaults("QEMU", product,
1433                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1434                         true, SMBIOS_ENTRY_POINT_30);
1435 
1436     smbios_get_tables(MACHINE(vms), NULL, 0, &smbios_tables, &smbios_tables_len,
1437                       &smbios_anchor, &smbios_anchor_len);
1438 
1439     if (smbios_anchor) {
1440         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1441                         smbios_tables, smbios_tables_len);
1442         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1443                         smbios_anchor, smbios_anchor_len);
1444     }
1445 }
1446 
1447 static
1448 void virt_machine_done(Notifier *notifier, void *data)
1449 {
1450     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1451                                          machine_done);
1452     MachineState *ms = MACHINE(vms);
1453     ARMCPU *cpu = ARM_CPU(first_cpu);
1454     struct arm_boot_info *info = &vms->bootinfo;
1455     AddressSpace *as = arm_boot_address_space(cpu, info);
1456 
1457     /*
1458      * If the user provided a dtb, we assume the dynamic sysbus nodes
1459      * already are integrated there. This corresponds to a use case where
1460      * the dynamic sysbus nodes are complex and their generation is not yet
1461      * supported. In that case the user can take charge of the guest dt
1462      * while qemu takes charge of the qom stuff.
1463      */
1464     if (info->dtb_filename == NULL) {
1465         platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1466                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1467                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1468                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1469     }
1470     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1471         exit(1);
1472     }
1473 
1474     virt_acpi_setup(vms);
1475     virt_build_smbios(vms);
1476 }
1477 
1478 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1479 {
1480     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1481     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1482 
1483     if (!vmc->disallow_affinity_adjustment) {
1484         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1485          * GIC's target-list limitations. 32-bit KVM hosts currently
1486          * always create clusters of 4 CPUs, but that is expected to
1487          * change when they gain support for gicv3. When KVM is enabled
1488          * it will override the changes we make here, therefore our
1489          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1490          * and to improve SGI efficiency.
1491          */
1492         if (vms->gic_version == VIRT_GIC_VERSION_3) {
1493             clustersz = GICV3_TARGETLIST_BITS;
1494         } else {
1495             clustersz = GIC_TARGETLIST_BITS;
1496         }
1497     }
1498     return arm_cpu_mp_affinity(idx, clustersz);
1499 }
1500 
1501 static void virt_set_memmap(VirtMachineState *vms)
1502 {
1503     MachineState *ms = MACHINE(vms);
1504     hwaddr base, device_memory_base, device_memory_size;
1505     int i;
1506 
1507     vms->memmap = extended_memmap;
1508 
1509     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1510         vms->memmap[i] = base_memmap[i];
1511     }
1512 
1513     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1514         error_report("unsupported number of memory slots: %"PRIu64,
1515                      ms->ram_slots);
1516         exit(EXIT_FAILURE);
1517     }
1518 
1519     /*
1520      * We compute the base of the high IO region depending on the
1521      * amount of initial and device memory. The device memory start/size
1522      * is aligned on 1GiB. We never put the high IO region below 256GiB
1523      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1524      * The device region size assumes 1GiB page max alignment per slot.
1525      */
1526     device_memory_base =
1527         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1528     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1529 
1530     /* Base address of the high IO region */
1531     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1532     if (base < device_memory_base) {
1533         error_report("maxmem/slots too huge");
1534         exit(EXIT_FAILURE);
1535     }
1536     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1537         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1538     }
1539 
1540     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1541         hwaddr size = extended_memmap[i].size;
1542 
1543         base = ROUND_UP(base, size);
1544         vms->memmap[i].base = base;
1545         vms->memmap[i].size = size;
1546         base += size;
1547     }
1548     vms->highest_gpa = base - 1;
1549     if (device_memory_size > 0) {
1550         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1551         ms->device_memory->base = device_memory_base;
1552         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1553                            "device-memory", device_memory_size);
1554     }
1555 }
1556 
1557 /*
1558  * finalize_gic_version - Determines the final gic_version
1559  * according to the gic-version property
1560  *
1561  * Default GIC type is v2
1562  */
1563 static void finalize_gic_version(VirtMachineState *vms)
1564 {
1565     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1566 
1567     if (kvm_enabled()) {
1568         int probe_bitmap;
1569 
1570         if (!kvm_irqchip_in_kernel()) {
1571             switch (vms->gic_version) {
1572             case VIRT_GIC_VERSION_HOST:
1573                 warn_report(
1574                     "gic-version=host not relevant with kernel-irqchip=off "
1575                      "as only userspace GICv2 is supported. Using v2 ...");
1576                 return;
1577             case VIRT_GIC_VERSION_MAX:
1578             case VIRT_GIC_VERSION_NOSEL:
1579                 vms->gic_version = VIRT_GIC_VERSION_2;
1580                 return;
1581             case VIRT_GIC_VERSION_2:
1582                 return;
1583             case VIRT_GIC_VERSION_3:
1584                 error_report(
1585                     "gic-version=3 is not supported with kernel-irqchip=off");
1586                 exit(1);
1587             }
1588         }
1589 
1590         probe_bitmap = kvm_arm_vgic_probe();
1591         if (!probe_bitmap) {
1592             error_report("Unable to determine GIC version supported by host");
1593             exit(1);
1594         }
1595 
1596         switch (vms->gic_version) {
1597         case VIRT_GIC_VERSION_HOST:
1598         case VIRT_GIC_VERSION_MAX:
1599             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1600                 vms->gic_version = VIRT_GIC_VERSION_3;
1601             } else {
1602                 vms->gic_version = VIRT_GIC_VERSION_2;
1603             }
1604             return;
1605         case VIRT_GIC_VERSION_NOSEL:
1606             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1607                 vms->gic_version = VIRT_GIC_VERSION_2;
1608             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1609                 /*
1610                  * in case the host does not support v2 in-kernel emulation or
1611                  * the end-user requested more than 8 VCPUs we now default
1612                  * to v3. In any case defaulting to v2 would be broken.
1613                  */
1614                 vms->gic_version = VIRT_GIC_VERSION_3;
1615             } else if (max_cpus > GIC_NCPU) {
1616                 error_report("host only supports in-kernel GICv2 emulation "
1617                              "but more than 8 vcpus are requested");
1618                 exit(1);
1619             }
1620             break;
1621         case VIRT_GIC_VERSION_2:
1622         case VIRT_GIC_VERSION_3:
1623             break;
1624         }
1625 
1626         /* Check chosen version is effectively supported by the host */
1627         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1628             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1629             error_report("host does not support in-kernel GICv2 emulation");
1630             exit(1);
1631         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1632                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1633             error_report("host does not support in-kernel GICv3 emulation");
1634             exit(1);
1635         }
1636         return;
1637     }
1638 
1639     /* TCG mode */
1640     switch (vms->gic_version) {
1641     case VIRT_GIC_VERSION_NOSEL:
1642         vms->gic_version = VIRT_GIC_VERSION_2;
1643         break;
1644     case VIRT_GIC_VERSION_MAX:
1645         vms->gic_version = VIRT_GIC_VERSION_3;
1646         break;
1647     case VIRT_GIC_VERSION_HOST:
1648         error_report("gic-version=host requires KVM");
1649         exit(1);
1650     case VIRT_GIC_VERSION_2:
1651     case VIRT_GIC_VERSION_3:
1652         break;
1653     }
1654 }
1655 
1656 static void machvirt_init(MachineState *machine)
1657 {
1658     VirtMachineState *vms = VIRT_MACHINE(machine);
1659     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1660     MachineClass *mc = MACHINE_GET_CLASS(machine);
1661     const CPUArchIdList *possible_cpus;
1662     MemoryRegion *sysmem = get_system_memory();
1663     MemoryRegion *secure_sysmem = NULL;
1664     int n, virt_max_cpus;
1665     bool firmware_loaded;
1666     bool aarch64 = true;
1667     bool has_ged = !vmc->no_ged;
1668     unsigned int smp_cpus = machine->smp.cpus;
1669     unsigned int max_cpus = machine->smp.max_cpus;
1670 
1671     /*
1672      * In accelerated mode, the memory map is computed earlier in kvm_type()
1673      * to create a VM with the right number of IPA bits.
1674      */
1675     if (!vms->memmap) {
1676         virt_set_memmap(vms);
1677     }
1678 
1679     /* We can probe only here because during property set
1680      * KVM is not available yet
1681      */
1682     finalize_gic_version(vms);
1683 
1684     if (!cpu_type_valid(machine->cpu_type)) {
1685         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1686         exit(1);
1687     }
1688 
1689     if (vms->secure) {
1690         if (kvm_enabled()) {
1691             error_report("mach-virt: KVM does not support Security extensions");
1692             exit(1);
1693         }
1694 
1695         /*
1696          * The Secure view of the world is the same as the NonSecure,
1697          * but with a few extra devices. Create it as a container region
1698          * containing the system memory at low priority; any secure-only
1699          * devices go in at higher priority and take precedence.
1700          */
1701         secure_sysmem = g_new(MemoryRegion, 1);
1702         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1703                            UINT64_MAX);
1704         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1705     }
1706 
1707     firmware_loaded = virt_firmware_init(vms, sysmem,
1708                                          secure_sysmem ?: sysmem);
1709 
1710     /* If we have an EL3 boot ROM then the assumption is that it will
1711      * implement PSCI itself, so disable QEMU's internal implementation
1712      * so it doesn't get in the way. Instead of starting secondary
1713      * CPUs in PSCI powerdown state we will start them all running and
1714      * let the boot ROM sort them out.
1715      * The usual case is that we do use QEMU's PSCI implementation;
1716      * if the guest has EL2 then we will use SMC as the conduit,
1717      * and otherwise we will use HVC (for backwards compatibility and
1718      * because if we're using KVM then we must use HVC).
1719      */
1720     if (vms->secure && firmware_loaded) {
1721         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1722     } else if (vms->virt) {
1723         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1724     } else {
1725         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1726     }
1727 
1728     /* The maximum number of CPUs depends on the GIC version, or on how
1729      * many redistributors we can fit into the memory map.
1730      */
1731     if (vms->gic_version == VIRT_GIC_VERSION_3) {
1732         virt_max_cpus =
1733             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1734         virt_max_cpus +=
1735             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1736     } else {
1737         virt_max_cpus = GIC_NCPU;
1738     }
1739 
1740     if (max_cpus > virt_max_cpus) {
1741         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1742                      "supported by machine 'mach-virt' (%d)",
1743                      max_cpus, virt_max_cpus);
1744         exit(1);
1745     }
1746 
1747     vms->smp_cpus = smp_cpus;
1748 
1749     if (vms->virt && kvm_enabled()) {
1750         error_report("mach-virt: KVM does not support providing "
1751                      "Virtualization extensions to the guest CPU");
1752         exit(1);
1753     }
1754 
1755     create_fdt(vms);
1756 
1757     possible_cpus = mc->possible_cpu_arch_ids(machine);
1758     for (n = 0; n < possible_cpus->len; n++) {
1759         Object *cpuobj;
1760         CPUState *cs;
1761 
1762         if (n >= smp_cpus) {
1763             break;
1764         }
1765 
1766         cpuobj = object_new(possible_cpus->cpus[n].type);
1767         object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
1768                                 "mp-affinity", NULL);
1769 
1770         cs = CPU(cpuobj);
1771         cs->cpu_index = n;
1772 
1773         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1774                           &error_fatal);
1775 
1776         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1777 
1778         if (!vms->secure) {
1779             object_property_set_bool(cpuobj, false, "has_el3", NULL);
1780         }
1781 
1782         if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
1783             object_property_set_bool(cpuobj, false, "has_el2", NULL);
1784         }
1785 
1786         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1787             object_property_set_int(cpuobj, vms->psci_conduit,
1788                                     "psci-conduit", NULL);
1789 
1790             /* Secondary CPUs start in PSCI powered-down state */
1791             if (n > 0) {
1792                 object_property_set_bool(cpuobj, true,
1793                                          "start-powered-off", NULL);
1794             }
1795         }
1796 
1797         if (vmc->kvm_no_adjvtime &&
1798             object_property_find(cpuobj, "kvm-no-adjvtime", NULL)) {
1799             object_property_set_bool(cpuobj, true, "kvm-no-adjvtime", NULL);
1800         }
1801 
1802         if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1803             object_property_set_bool(cpuobj, false, "pmu", NULL);
1804         }
1805 
1806         if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1807             object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
1808                                     "reset-cbar", &error_abort);
1809         }
1810 
1811         object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1812                                  &error_abort);
1813         if (vms->secure) {
1814             object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1815                                      "secure-memory", &error_abort);
1816         }
1817 
1818         object_property_set_bool(cpuobj, true, "realized", &error_fatal);
1819         object_unref(cpuobj);
1820     }
1821     fdt_add_timer_nodes(vms);
1822     fdt_add_cpu_nodes(vms);
1823 
1824    if (!kvm_enabled()) {
1825         ARMCPU *cpu = ARM_CPU(first_cpu);
1826         bool aarch64 = object_property_get_bool(OBJECT(cpu), "aarch64", NULL);
1827 
1828         if (aarch64 && vms->highmem) {
1829             int requested_pa_size, pamax = arm_pamax(cpu);
1830 
1831             requested_pa_size = 64 - clz64(vms->highest_gpa);
1832             if (pamax < requested_pa_size) {
1833                 error_report("VCPU supports less PA bits (%d) than requested "
1834                             "by the memory map (%d)", pamax, requested_pa_size);
1835                 exit(1);
1836             }
1837         }
1838     }
1839 
1840     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
1841                                 machine->ram);
1842     if (machine->device_memory) {
1843         memory_region_add_subregion(sysmem, machine->device_memory->base,
1844                                     &machine->device_memory->mr);
1845     }
1846 
1847     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
1848 
1849     create_gic(vms);
1850 
1851     fdt_add_pmu_nodes(vms);
1852 
1853     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
1854 
1855     if (vms->secure) {
1856         create_secure_ram(vms, secure_sysmem);
1857         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
1858     }
1859 
1860     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
1861 
1862     create_rtc(vms);
1863 
1864     create_pcie(vms);
1865 
1866     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
1867         vms->acpi_dev = create_acpi_ged(vms);
1868     } else {
1869         create_gpio(vms);
1870     }
1871 
1872      /* connect powerdown request */
1873      vms->powerdown_notifier.notify = virt_powerdown_req;
1874      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
1875 
1876     /* Create mmio transports, so the user can create virtio backends
1877      * (which will be automatically plugged in to the transports). If
1878      * no backend is created the transport will just sit harmlessly idle.
1879      */
1880     create_virtio_devices(vms);
1881 
1882     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1883     rom_set_fw(vms->fw_cfg);
1884 
1885     create_platform_bus(vms);
1886 
1887     vms->bootinfo.ram_size = machine->ram_size;
1888     vms->bootinfo.nb_cpus = smp_cpus;
1889     vms->bootinfo.board_id = -1;
1890     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1891     vms->bootinfo.get_dtb = machvirt_dtb;
1892     vms->bootinfo.skip_dtb_autoload = true;
1893     vms->bootinfo.firmware_loaded = firmware_loaded;
1894     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
1895 
1896     vms->machine_done.notify = virt_machine_done;
1897     qemu_add_machine_init_done_notifier(&vms->machine_done);
1898 }
1899 
1900 static bool virt_get_secure(Object *obj, Error **errp)
1901 {
1902     VirtMachineState *vms = VIRT_MACHINE(obj);
1903 
1904     return vms->secure;
1905 }
1906 
1907 static void virt_set_secure(Object *obj, bool value, Error **errp)
1908 {
1909     VirtMachineState *vms = VIRT_MACHINE(obj);
1910 
1911     vms->secure = value;
1912 }
1913 
1914 static bool virt_get_virt(Object *obj, Error **errp)
1915 {
1916     VirtMachineState *vms = VIRT_MACHINE(obj);
1917 
1918     return vms->virt;
1919 }
1920 
1921 static void virt_set_virt(Object *obj, bool value, Error **errp)
1922 {
1923     VirtMachineState *vms = VIRT_MACHINE(obj);
1924 
1925     vms->virt = value;
1926 }
1927 
1928 static bool virt_get_highmem(Object *obj, Error **errp)
1929 {
1930     VirtMachineState *vms = VIRT_MACHINE(obj);
1931 
1932     return vms->highmem;
1933 }
1934 
1935 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1936 {
1937     VirtMachineState *vms = VIRT_MACHINE(obj);
1938 
1939     vms->highmem = value;
1940 }
1941 
1942 static bool virt_get_its(Object *obj, Error **errp)
1943 {
1944     VirtMachineState *vms = VIRT_MACHINE(obj);
1945 
1946     return vms->its;
1947 }
1948 
1949 static void virt_set_its(Object *obj, bool value, Error **errp)
1950 {
1951     VirtMachineState *vms = VIRT_MACHINE(obj);
1952 
1953     vms->its = value;
1954 }
1955 
1956 bool virt_is_acpi_enabled(VirtMachineState *vms)
1957 {
1958     if (vms->acpi == ON_OFF_AUTO_OFF) {
1959         return false;
1960     }
1961     return true;
1962 }
1963 
1964 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
1965                           void *opaque, Error **errp)
1966 {
1967     VirtMachineState *vms = VIRT_MACHINE(obj);
1968     OnOffAuto acpi = vms->acpi;
1969 
1970     visit_type_OnOffAuto(v, name, &acpi, errp);
1971 }
1972 
1973 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
1974                           void *opaque, Error **errp)
1975 {
1976     VirtMachineState *vms = VIRT_MACHINE(obj);
1977 
1978     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
1979 }
1980 
1981 static char *virt_get_gic_version(Object *obj, Error **errp)
1982 {
1983     VirtMachineState *vms = VIRT_MACHINE(obj);
1984     const char *val = vms->gic_version == VIRT_GIC_VERSION_3 ? "3" : "2";
1985 
1986     return g_strdup(val);
1987 }
1988 
1989 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1990 {
1991     VirtMachineState *vms = VIRT_MACHINE(obj);
1992 
1993     if (!strcmp(value, "3")) {
1994         vms->gic_version = VIRT_GIC_VERSION_3;
1995     } else if (!strcmp(value, "2")) {
1996         vms->gic_version = VIRT_GIC_VERSION_2;
1997     } else if (!strcmp(value, "host")) {
1998         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
1999     } else if (!strcmp(value, "max")) {
2000         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2001     } else {
2002         error_setg(errp, "Invalid gic-version value");
2003         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2004     }
2005 }
2006 
2007 static char *virt_get_iommu(Object *obj, Error **errp)
2008 {
2009     VirtMachineState *vms = VIRT_MACHINE(obj);
2010 
2011     switch (vms->iommu) {
2012     case VIRT_IOMMU_NONE:
2013         return g_strdup("none");
2014     case VIRT_IOMMU_SMMUV3:
2015         return g_strdup("smmuv3");
2016     default:
2017         g_assert_not_reached();
2018     }
2019 }
2020 
2021 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2022 {
2023     VirtMachineState *vms = VIRT_MACHINE(obj);
2024 
2025     if (!strcmp(value, "smmuv3")) {
2026         vms->iommu = VIRT_IOMMU_SMMUV3;
2027     } else if (!strcmp(value, "none")) {
2028         vms->iommu = VIRT_IOMMU_NONE;
2029     } else {
2030         error_setg(errp, "Invalid iommu value");
2031         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2032     }
2033 }
2034 
2035 static CpuInstanceProperties
2036 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2037 {
2038     MachineClass *mc = MACHINE_GET_CLASS(ms);
2039     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2040 
2041     assert(cpu_index < possible_cpus->len);
2042     return possible_cpus->cpus[cpu_index].props;
2043 }
2044 
2045 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2046 {
2047     return idx % ms->numa_state->num_nodes;
2048 }
2049 
2050 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2051 {
2052     int n;
2053     unsigned int max_cpus = ms->smp.max_cpus;
2054     VirtMachineState *vms = VIRT_MACHINE(ms);
2055 
2056     if (ms->possible_cpus) {
2057         assert(ms->possible_cpus->len == max_cpus);
2058         return ms->possible_cpus;
2059     }
2060 
2061     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2062                                   sizeof(CPUArchId) * max_cpus);
2063     ms->possible_cpus->len = max_cpus;
2064     for (n = 0; n < ms->possible_cpus->len; n++) {
2065         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2066         ms->possible_cpus->cpus[n].arch_id =
2067             virt_cpu_mp_affinity(vms, n);
2068         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2069         ms->possible_cpus->cpus[n].props.thread_id = n;
2070     }
2071     return ms->possible_cpus;
2072 }
2073 
2074 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2075                                  Error **errp)
2076 {
2077     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2078     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2079 
2080     if (is_nvdimm) {
2081         error_setg(errp, "nvdimm is not yet supported");
2082         return;
2083     }
2084 
2085     if (!vms->acpi_dev) {
2086         error_setg(errp,
2087                    "memory hotplug is not enabled: missing acpi-ged device");
2088         return;
2089     }
2090 
2091     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2092 }
2093 
2094 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2095                              DeviceState *dev, Error **errp)
2096 {
2097     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2098     Error *local_err = NULL;
2099 
2100     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms), &local_err);
2101     if (local_err) {
2102         goto out;
2103     }
2104 
2105     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2106                          dev, &error_abort);
2107 
2108 out:
2109     error_propagate(errp, local_err);
2110 }
2111 
2112 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2113                                             DeviceState *dev, Error **errp)
2114 {
2115     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2116         virt_memory_pre_plug(hotplug_dev, dev, errp);
2117     }
2118 }
2119 
2120 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2121                                         DeviceState *dev, Error **errp)
2122 {
2123     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2124 
2125     if (vms->platform_bus_dev) {
2126         if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
2127             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2128                                      SYS_BUS_DEVICE(dev));
2129         }
2130     }
2131     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2132         virt_memory_plug(hotplug_dev, dev, errp);
2133     }
2134     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2135         PCIDevice *pdev = PCI_DEVICE(dev);
2136 
2137         vms->iommu = VIRT_IOMMU_VIRTIO;
2138         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2139         create_virtio_iommu_dt_bindings(vms);
2140     }
2141 }
2142 
2143 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2144                                           DeviceState *dev, Error **errp)
2145 {
2146     error_setg(errp, "device unplug request for unsupported device"
2147                " type: %s", object_get_typename(OBJECT(dev)));
2148 }
2149 
2150 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2151                                                         DeviceState *dev)
2152 {
2153     if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE) ||
2154        (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
2155         return HOTPLUG_HANDLER(machine);
2156     }
2157     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2158         VirtMachineState *vms = VIRT_MACHINE(machine);
2159 
2160         if (!vms->bootinfo.firmware_loaded || !virt_is_acpi_enabled(vms)) {
2161             return HOTPLUG_HANDLER(machine);
2162         }
2163     }
2164     return NULL;
2165 }
2166 
2167 /*
2168  * for arm64 kvm_type [7-0] encodes the requested number of bits
2169  * in the IPA address space
2170  */
2171 static int virt_kvm_type(MachineState *ms, const char *type_str)
2172 {
2173     VirtMachineState *vms = VIRT_MACHINE(ms);
2174     int max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms);
2175     int requested_pa_size;
2176 
2177     /* we freeze the memory map to compute the highest gpa */
2178     virt_set_memmap(vms);
2179 
2180     requested_pa_size = 64 - clz64(vms->highest_gpa);
2181 
2182     if (requested_pa_size > max_vm_pa_size) {
2183         error_report("-m and ,maxmem option values "
2184                      "require an IPA range (%d bits) larger than "
2185                      "the one supported by the host (%d bits)",
2186                      requested_pa_size, max_vm_pa_size);
2187        exit(1);
2188     }
2189     /*
2190      * By default we return 0 which corresponds to an implicit legacy
2191      * 40b IPA setting. Otherwise we return the actual requested PA
2192      * logsize
2193      */
2194     return requested_pa_size > 40 ? requested_pa_size : 0;
2195 }
2196 
2197 static void virt_machine_class_init(ObjectClass *oc, void *data)
2198 {
2199     MachineClass *mc = MACHINE_CLASS(oc);
2200     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2201 
2202     mc->init = machvirt_init;
2203     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2204      * The value may be reduced later when we have more information about the
2205      * configuration of the particular instance.
2206      */
2207     mc->max_cpus = 512;
2208     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2209     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2210     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2211     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2212     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2213     mc->block_default_type = IF_VIRTIO;
2214     mc->no_cdrom = 1;
2215     mc->pci_allow_0_address = true;
2216     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2217     mc->minimum_page_bits = 12;
2218     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2219     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2220     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2221     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2222     mc->kvm_type = virt_kvm_type;
2223     assert(!mc->get_hotplug_handler);
2224     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2225     hc->pre_plug = virt_machine_device_pre_plug_cb;
2226     hc->plug = virt_machine_device_plug_cb;
2227     hc->unplug_request = virt_machine_device_unplug_request_cb;
2228     mc->numa_mem_supported = true;
2229     mc->auto_enable_numa_with_memhp = true;
2230     mc->default_ram_id = "mach-virt.ram";
2231 
2232     object_class_property_add(oc, "acpi", "OnOffAuto",
2233         virt_get_acpi, virt_set_acpi,
2234         NULL, NULL, &error_abort);
2235     object_class_property_set_description(oc, "acpi",
2236         "Enable ACPI", &error_abort);
2237 }
2238 
2239 static void virt_instance_init(Object *obj)
2240 {
2241     VirtMachineState *vms = VIRT_MACHINE(obj);
2242     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2243 
2244     /* EL3 is disabled by default on virt: this makes us consistent
2245      * between KVM and TCG for this board, and it also allows us to
2246      * boot UEFI blobs which assume no TrustZone support.
2247      */
2248     vms->secure = false;
2249     object_property_add_bool(obj, "secure", virt_get_secure,
2250                              virt_set_secure, NULL);
2251     object_property_set_description(obj, "secure",
2252                                     "Set on/off to enable/disable the ARM "
2253                                     "Security Extensions (TrustZone)",
2254                                     NULL);
2255 
2256     /* EL2 is also disabled by default, for similar reasons */
2257     vms->virt = false;
2258     object_property_add_bool(obj, "virtualization", virt_get_virt,
2259                              virt_set_virt, NULL);
2260     object_property_set_description(obj, "virtualization",
2261                                     "Set on/off to enable/disable emulating a "
2262                                     "guest CPU which implements the ARM "
2263                                     "Virtualization Extensions",
2264                                     NULL);
2265 
2266     /* High memory is enabled by default */
2267     vms->highmem = true;
2268     object_property_add_bool(obj, "highmem", virt_get_highmem,
2269                              virt_set_highmem, NULL);
2270     object_property_set_description(obj, "highmem",
2271                                     "Set on/off to enable/disable using "
2272                                     "physical address space above 32 bits",
2273                                     NULL);
2274     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
2275     object_property_add_str(obj, "gic-version", virt_get_gic_version,
2276                         virt_set_gic_version, NULL);
2277     object_property_set_description(obj, "gic-version",
2278                                     "Set GIC version. "
2279                                     "Valid values are 2, 3, host and max",
2280                                     NULL);
2281 
2282     vms->highmem_ecam = !vmc->no_highmem_ecam;
2283 
2284     if (vmc->no_its) {
2285         vms->its = false;
2286     } else {
2287         /* Default allows ITS instantiation */
2288         vms->its = true;
2289         object_property_add_bool(obj, "its", virt_get_its,
2290                                  virt_set_its, NULL);
2291         object_property_set_description(obj, "its",
2292                                         "Set on/off to enable/disable "
2293                                         "ITS instantiation",
2294                                         NULL);
2295     }
2296 
2297     /* Default disallows iommu instantiation */
2298     vms->iommu = VIRT_IOMMU_NONE;
2299     object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu, NULL);
2300     object_property_set_description(obj, "iommu",
2301                                     "Set the IOMMU type. "
2302                                     "Valid values are none and smmuv3",
2303                                     NULL);
2304 
2305     vms->irqmap = a15irqmap;
2306 
2307     virt_flash_create(vms);
2308 }
2309 
2310 static const TypeInfo virt_machine_info = {
2311     .name          = TYPE_VIRT_MACHINE,
2312     .parent        = TYPE_MACHINE,
2313     .abstract      = true,
2314     .instance_size = sizeof(VirtMachineState),
2315     .class_size    = sizeof(VirtMachineClass),
2316     .class_init    = virt_machine_class_init,
2317     .instance_init = virt_instance_init,
2318     .interfaces = (InterfaceInfo[]) {
2319          { TYPE_HOTPLUG_HANDLER },
2320          { }
2321     },
2322 };
2323 
2324 static void machvirt_machine_init(void)
2325 {
2326     type_register_static(&virt_machine_info);
2327 }
2328 type_init(machvirt_machine_init);
2329 
2330 static void virt_machine_5_0_options(MachineClass *mc)
2331 {
2332     static GlobalProperty compat[] = {
2333         { TYPE_TPM_TIS_SYSBUS, "ppi", "false" },
2334     };
2335 
2336     compat_props_add(mc->compat_props, compat, G_N_ELEMENTS(compat));
2337 }
2338 DEFINE_VIRT_MACHINE_AS_LATEST(5, 0)
2339 
2340 static void virt_machine_4_2_options(MachineClass *mc)
2341 {
2342     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2343 
2344     virt_machine_5_0_options(mc);
2345     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
2346     vmc->kvm_no_adjvtime = true;
2347 }
2348 DEFINE_VIRT_MACHINE(4, 2)
2349 
2350 static void virt_machine_4_1_options(MachineClass *mc)
2351 {
2352     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2353 
2354     virt_machine_4_2_options(mc);
2355     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2356     vmc->no_ged = true;
2357     mc->auto_enable_numa_with_memhp = false;
2358 }
2359 DEFINE_VIRT_MACHINE(4, 1)
2360 
2361 static void virt_machine_4_0_options(MachineClass *mc)
2362 {
2363     virt_machine_4_1_options(mc);
2364     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2365 }
2366 DEFINE_VIRT_MACHINE(4, 0)
2367 
2368 static void virt_machine_3_1_options(MachineClass *mc)
2369 {
2370     virt_machine_4_0_options(mc);
2371     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2372 }
2373 DEFINE_VIRT_MACHINE(3, 1)
2374 
2375 static void virt_machine_3_0_options(MachineClass *mc)
2376 {
2377     virt_machine_3_1_options(mc);
2378     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2379 }
2380 DEFINE_VIRT_MACHINE(3, 0)
2381 
2382 static void virt_machine_2_12_options(MachineClass *mc)
2383 {
2384     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2385 
2386     virt_machine_3_0_options(mc);
2387     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2388     vmc->no_highmem_ecam = true;
2389     mc->max_cpus = 255;
2390 }
2391 DEFINE_VIRT_MACHINE(2, 12)
2392 
2393 static void virt_machine_2_11_options(MachineClass *mc)
2394 {
2395     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2396 
2397     virt_machine_2_12_options(mc);
2398     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2399     vmc->smbios_old_sys_ver = true;
2400 }
2401 DEFINE_VIRT_MACHINE(2, 11)
2402 
2403 static void virt_machine_2_10_options(MachineClass *mc)
2404 {
2405     virt_machine_2_11_options(mc);
2406     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2407     /* before 2.11 we never faulted accesses to bad addresses */
2408     mc->ignore_memory_transaction_failures = true;
2409 }
2410 DEFINE_VIRT_MACHINE(2, 10)
2411 
2412 static void virt_machine_2_9_options(MachineClass *mc)
2413 {
2414     virt_machine_2_10_options(mc);
2415     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2416 }
2417 DEFINE_VIRT_MACHINE(2, 9)
2418 
2419 static void virt_machine_2_8_options(MachineClass *mc)
2420 {
2421     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2422 
2423     virt_machine_2_9_options(mc);
2424     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
2425     /* For 2.8 and earlier we falsely claimed in the DT that
2426      * our timers were edge-triggered, not level-triggered.
2427      */
2428     vmc->claim_edge_triggered_timers = true;
2429 }
2430 DEFINE_VIRT_MACHINE(2, 8)
2431 
2432 static void virt_machine_2_7_options(MachineClass *mc)
2433 {
2434     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2435 
2436     virt_machine_2_8_options(mc);
2437     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2438     /* ITS was introduced with 2.8 */
2439     vmc->no_its = true;
2440     /* Stick with 1K pages for migration compatibility */
2441     mc->minimum_page_bits = 0;
2442 }
2443 DEFINE_VIRT_MACHINE(2, 7)
2444 
2445 static void virt_machine_2_6_options(MachineClass *mc)
2446 {
2447     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2448 
2449     virt_machine_2_7_options(mc);
2450     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
2451     vmc->disallow_affinity_adjustment = true;
2452     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2453     vmc->no_pmu = true;
2454 }
2455 DEFINE_VIRT_MACHINE(2, 6)
2456