xref: /openbmc/qemu/hw/arm/virt.c (revision 3710586c)
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_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
154     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
155     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
156     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
157     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
158     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
159     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
160     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
161     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
162     /* Actual RAM size depends on initial RAM and device memory settings */
163     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
164 };
165 
166 /*
167  * Highmem IO Regions: This memory map is floating, located after the RAM.
168  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
169  * top of the RAM, so that its base get the same alignment as the size,
170  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
171  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
172  * Note the extended_memmap is sized so that it eventually also includes the
173  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
174  * index of base_memmap).
175  */
176 static MemMapEntry extended_memmap[] = {
177     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
178     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
179     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
180     /* Second PCIe window */
181     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
182 };
183 
184 static const int a15irqmap[] = {
185     [VIRT_UART] = 1,
186     [VIRT_RTC] = 2,
187     [VIRT_PCIE] = 3, /* ... to 6 */
188     [VIRT_GPIO] = 7,
189     [VIRT_SECURE_UART] = 8,
190     [VIRT_ACPI_GED] = 9,
191     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
192     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
193     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
194     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
195 };
196 
197 static const char *valid_cpus[] = {
198     ARM_CPU_TYPE_NAME("cortex-a7"),
199     ARM_CPU_TYPE_NAME("cortex-a15"),
200     ARM_CPU_TYPE_NAME("cortex-a53"),
201     ARM_CPU_TYPE_NAME("cortex-a57"),
202     ARM_CPU_TYPE_NAME("cortex-a72"),
203     ARM_CPU_TYPE_NAME("host"),
204     ARM_CPU_TYPE_NAME("max"),
205 };
206 
207 static bool cpu_type_valid(const char *cpu)
208 {
209     int i;
210 
211     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
212         if (strcmp(cpu, valid_cpus[i]) == 0) {
213             return true;
214         }
215     }
216     return false;
217 }
218 
219 static void create_kaslr_seed(VirtMachineState *vms, const char *node)
220 {
221     uint64_t seed;
222 
223     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
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     ARMCPU *armcpu = ARM_CPU(first_cpu);
526     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
527 
528     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
529         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
530         return;
531     }
532 
533     if (vms->gic_version == VIRT_GIC_VERSION_2) {
534         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
535                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
536                              (1 << vms->smp_cpus) - 1);
537     }
538 
539     qemu_fdt_add_subnode(vms->fdt, "/pmu");
540     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
541         const char compat[] = "arm,armv8-pmuv3";
542         qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
543                          compat, sizeof(compat));
544         qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
545                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
546     }
547 }
548 
549 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
550 {
551     DeviceState *dev;
552     MachineState *ms = MACHINE(vms);
553     int irq = vms->irqmap[VIRT_ACPI_GED];
554     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
555 
556     if (ms->ram_slots) {
557         event |= ACPI_GED_MEM_HOTPLUG_EVT;
558     }
559 
560     if (ms->nvdimms_state->is_enabled) {
561         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
562     }
563 
564     dev = qdev_new(TYPE_ACPI_GED);
565     qdev_prop_set_uint32(dev, "ged-event", event);
566 
567     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
568     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
569     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
570 
571     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
572 
573     return dev;
574 }
575 
576 static void create_its(VirtMachineState *vms)
577 {
578     const char *itsclass = its_class_name();
579     DeviceState *dev;
580 
581     if (!itsclass) {
582         /* Do nothing if not supported */
583         return;
584     }
585 
586     dev = qdev_new(itsclass);
587 
588     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
589                              &error_abort);
590     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
591     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
592 
593     fdt_add_its_gic_node(vms);
594     vms->msi_controller = VIRT_MSI_CTRL_ITS;
595 }
596 
597 static void create_v2m(VirtMachineState *vms)
598 {
599     int i;
600     int irq = vms->irqmap[VIRT_GIC_V2M];
601     DeviceState *dev;
602 
603     dev = qdev_new("arm-gicv2m");
604     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
605     qdev_prop_set_uint32(dev, "base-spi", irq);
606     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
607     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
608 
609     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
610         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
611                            qdev_get_gpio_in(vms->gic, irq + i));
612     }
613 
614     fdt_add_v2m_gic_node(vms);
615     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
616 }
617 
618 static void create_gic(VirtMachineState *vms)
619 {
620     MachineState *ms = MACHINE(vms);
621     /* We create a standalone GIC */
622     SysBusDevice *gicbusdev;
623     const char *gictype;
624     int type = vms->gic_version, i;
625     unsigned int smp_cpus = ms->smp.cpus;
626     uint32_t nb_redist_regions = 0;
627 
628     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
629 
630     vms->gic = qdev_new(gictype);
631     qdev_prop_set_uint32(vms->gic, "revision", type);
632     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
633     /* Note that the num-irq property counts both internal and external
634      * interrupts; there are always 32 of the former (mandated by GIC spec).
635      */
636     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
637     if (!kvm_irqchip_in_kernel()) {
638         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
639     }
640 
641     if (type == 3) {
642         uint32_t redist0_capacity =
643                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
644         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
645 
646         nb_redist_regions = virt_gicv3_redist_region_count(vms);
647 
648         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
649                              nb_redist_regions);
650         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
651 
652         if (nb_redist_regions == 2) {
653             uint32_t redist1_capacity =
654                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
655 
656             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
657                 MIN(smp_cpus - redist0_count, redist1_capacity));
658         }
659     } else {
660         if (!kvm_irqchip_in_kernel()) {
661             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
662                               vms->virt);
663         }
664     }
665     gicbusdev = SYS_BUS_DEVICE(vms->gic);
666     sysbus_realize_and_unref(gicbusdev, &error_fatal);
667     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
668     if (type == 3) {
669         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
670         if (nb_redist_regions == 2) {
671             sysbus_mmio_map(gicbusdev, 2,
672                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
673         }
674     } else {
675         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
676         if (vms->virt) {
677             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
678             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
679         }
680     }
681 
682     /* Wire the outputs from each CPU's generic timer and the GICv3
683      * maintenance interrupt signal to the appropriate GIC PPI inputs,
684      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
685      */
686     for (i = 0; i < smp_cpus; i++) {
687         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
688         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
689         int irq;
690         /* Mapping from the output timer irq lines from the CPU to the
691          * GIC PPI inputs we use for the virt board.
692          */
693         const int timer_irq[] = {
694             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
695             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
696             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
697             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
698         };
699 
700         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
701             qdev_connect_gpio_out(cpudev, irq,
702                                   qdev_get_gpio_in(vms->gic,
703                                                    ppibase + timer_irq[irq]));
704         }
705 
706         if (type == 3) {
707             qemu_irq irq = qdev_get_gpio_in(vms->gic,
708                                             ppibase + ARCH_GIC_MAINT_IRQ);
709             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
710                                         0, irq);
711         } else if (vms->virt) {
712             qemu_irq irq = qdev_get_gpio_in(vms->gic,
713                                             ppibase + ARCH_GIC_MAINT_IRQ);
714             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
715         }
716 
717         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
718                                     qdev_get_gpio_in(vms->gic, ppibase
719                                                      + VIRTUAL_PMU_IRQ));
720 
721         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
722         sysbus_connect_irq(gicbusdev, i + smp_cpus,
723                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
724         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
725                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
726         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
727                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
728     }
729 
730     fdt_add_gic_node(vms);
731 
732     if (type == 3 && vms->its) {
733         create_its(vms);
734     } else if (type == 2) {
735         create_v2m(vms);
736     }
737 }
738 
739 static void create_uart(const VirtMachineState *vms, int uart,
740                         MemoryRegion *mem, Chardev *chr)
741 {
742     char *nodename;
743     hwaddr base = vms->memmap[uart].base;
744     hwaddr size = vms->memmap[uart].size;
745     int irq = vms->irqmap[uart];
746     const char compat[] = "arm,pl011\0arm,primecell";
747     const char clocknames[] = "uartclk\0apb_pclk";
748     DeviceState *dev = qdev_new(TYPE_PL011);
749     SysBusDevice *s = SYS_BUS_DEVICE(dev);
750 
751     qdev_prop_set_chr(dev, "chardev", chr);
752     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
753     memory_region_add_subregion(mem, base,
754                                 sysbus_mmio_get_region(s, 0));
755     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
756 
757     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
758     qemu_fdt_add_subnode(vms->fdt, nodename);
759     /* Note that we can't use setprop_string because of the embedded NUL */
760     qemu_fdt_setprop(vms->fdt, nodename, "compatible",
761                          compat, sizeof(compat));
762     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
763                                      2, base, 2, size);
764     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
765                                GIC_FDT_IRQ_TYPE_SPI, irq,
766                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
767     qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
768                                vms->clock_phandle, vms->clock_phandle);
769     qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
770                          clocknames, sizeof(clocknames));
771 
772     if (uart == VIRT_UART) {
773         qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
774     } else {
775         /* Mark as not usable by the normal world */
776         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
777         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
778 
779         qemu_fdt_setprop_string(vms->fdt, "/secure-chosen", "stdout-path",
780                                 nodename);
781     }
782 
783     g_free(nodename);
784 }
785 
786 static void create_rtc(const VirtMachineState *vms)
787 {
788     char *nodename;
789     hwaddr base = vms->memmap[VIRT_RTC].base;
790     hwaddr size = vms->memmap[VIRT_RTC].size;
791     int irq = vms->irqmap[VIRT_RTC];
792     const char compat[] = "arm,pl031\0arm,primecell";
793 
794     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
795 
796     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
797     qemu_fdt_add_subnode(vms->fdt, nodename);
798     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
799     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
800                                  2, base, 2, size);
801     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
802                            GIC_FDT_IRQ_TYPE_SPI, irq,
803                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
804     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
805     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
806     g_free(nodename);
807 }
808 
809 static DeviceState *gpio_key_dev;
810 static void virt_powerdown_req(Notifier *n, void *opaque)
811 {
812     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
813 
814     if (s->acpi_dev) {
815         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
816     } else {
817         /* use gpio Pin 3 for power button event */
818         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
819     }
820 }
821 
822 static void create_gpio(const VirtMachineState *vms)
823 {
824     char *nodename;
825     DeviceState *pl061_dev;
826     hwaddr base = vms->memmap[VIRT_GPIO].base;
827     hwaddr size = vms->memmap[VIRT_GPIO].size;
828     int irq = vms->irqmap[VIRT_GPIO];
829     const char compat[] = "arm,pl061\0arm,primecell";
830 
831     pl061_dev = sysbus_create_simple("pl061", base,
832                                      qdev_get_gpio_in(vms->gic, irq));
833 
834     uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
835     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
836     qemu_fdt_add_subnode(vms->fdt, nodename);
837     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
838                                  2, base, 2, size);
839     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
840     qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
841     qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
842     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
843                            GIC_FDT_IRQ_TYPE_SPI, irq,
844                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
845     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
846     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
847     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
848 
849     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
850                                         qdev_get_gpio_in(pl061_dev, 3));
851     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
852     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
853     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
854     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
855 
856     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
857     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
858                             "label", "GPIO Key Poweroff");
859     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
860                           KEY_POWER);
861     qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
862                            "gpios", phandle, 3, 0);
863     g_free(nodename);
864 }
865 
866 static void create_virtio_devices(const VirtMachineState *vms)
867 {
868     int i;
869     hwaddr size = vms->memmap[VIRT_MMIO].size;
870 
871     /* We create the transports in forwards order. Since qbus_realize()
872      * prepends (not appends) new child buses, the incrementing loop below will
873      * create a list of virtio-mmio buses with decreasing base addresses.
874      *
875      * When a -device option is processed from the command line,
876      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
877      * order. The upshot is that -device options in increasing command line
878      * order are mapped to virtio-mmio buses with decreasing base addresses.
879      *
880      * When this code was originally written, that arrangement ensured that the
881      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
882      * the first -device on the command line. (The end-to-end order is a
883      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
884      * guest kernel's name-to-address assignment strategy.)
885      *
886      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
887      * the message, if not necessarily the code, of commit 70161ff336.
888      * Therefore the loop now establishes the inverse of the original intent.
889      *
890      * Unfortunately, we can't counteract the kernel change by reversing the
891      * loop; it would break existing command lines.
892      *
893      * In any case, the kernel makes no guarantee about the stability of
894      * enumeration order of virtio devices (as demonstrated by it changing
895      * between kernel versions). For reliable and stable identification
896      * of disks users must use UUIDs or similar mechanisms.
897      */
898     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
899         int irq = vms->irqmap[VIRT_MMIO] + i;
900         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
901 
902         sysbus_create_simple("virtio-mmio", base,
903                              qdev_get_gpio_in(vms->gic, irq));
904     }
905 
906     /* We add dtb nodes in reverse order so that they appear in the finished
907      * device tree lowest address first.
908      *
909      * Note that this mapping is independent of the loop above. The previous
910      * loop influences virtio device to virtio transport assignment, whereas
911      * this loop controls how virtio transports are laid out in the dtb.
912      */
913     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
914         char *nodename;
915         int irq = vms->irqmap[VIRT_MMIO] + i;
916         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
917 
918         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
919         qemu_fdt_add_subnode(vms->fdt, nodename);
920         qemu_fdt_setprop_string(vms->fdt, nodename,
921                                 "compatible", "virtio,mmio");
922         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
923                                      2, base, 2, size);
924         qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
925                                GIC_FDT_IRQ_TYPE_SPI, irq,
926                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
927         qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
928         g_free(nodename);
929     }
930 }
931 
932 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
933 
934 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
935                                         const char *name,
936                                         const char *alias_prop_name)
937 {
938     /*
939      * Create a single flash device.  We use the same parameters as
940      * the flash devices on the Versatile Express board.
941      */
942     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
943 
944     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
945     qdev_prop_set_uint8(dev, "width", 4);
946     qdev_prop_set_uint8(dev, "device-width", 2);
947     qdev_prop_set_bit(dev, "big-endian", false);
948     qdev_prop_set_uint16(dev, "id0", 0x89);
949     qdev_prop_set_uint16(dev, "id1", 0x18);
950     qdev_prop_set_uint16(dev, "id2", 0x00);
951     qdev_prop_set_uint16(dev, "id3", 0x00);
952     qdev_prop_set_string(dev, "name", name);
953     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
954     object_property_add_alias(OBJECT(vms), alias_prop_name,
955                               OBJECT(dev), "drive");
956     return PFLASH_CFI01(dev);
957 }
958 
959 static void virt_flash_create(VirtMachineState *vms)
960 {
961     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
962     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
963 }
964 
965 static void virt_flash_map1(PFlashCFI01 *flash,
966                             hwaddr base, hwaddr size,
967                             MemoryRegion *sysmem)
968 {
969     DeviceState *dev = DEVICE(flash);
970 
971     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
972     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
973     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
974     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
975 
976     memory_region_add_subregion(sysmem, base,
977                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
978                                                        0));
979 }
980 
981 static void virt_flash_map(VirtMachineState *vms,
982                            MemoryRegion *sysmem,
983                            MemoryRegion *secure_sysmem)
984 {
985     /*
986      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
987      * sysmem is the system memory space. secure_sysmem is the secure view
988      * of the system, and the first flash device should be made visible only
989      * there. The second flash device is visible to both secure and nonsecure.
990      * If sysmem == secure_sysmem this means there is no separate Secure
991      * address space and both flash devices are generally visible.
992      */
993     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
994     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
995 
996     virt_flash_map1(vms->flash[0], flashbase, flashsize,
997                     secure_sysmem);
998     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
999                     sysmem);
1000 }
1001 
1002 static void virt_flash_fdt(VirtMachineState *vms,
1003                            MemoryRegion *sysmem,
1004                            MemoryRegion *secure_sysmem)
1005 {
1006     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1007     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1008     char *nodename;
1009 
1010     if (sysmem == secure_sysmem) {
1011         /* Report both flash devices as a single node in the DT */
1012         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1013         qemu_fdt_add_subnode(vms->fdt, nodename);
1014         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1015         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1016                                      2, flashbase, 2, flashsize,
1017                                      2, flashbase + flashsize, 2, flashsize);
1018         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1019         g_free(nodename);
1020     } else {
1021         /*
1022          * Report the devices as separate nodes so we can mark one as
1023          * only visible to the secure world.
1024          */
1025         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1026         qemu_fdt_add_subnode(vms->fdt, nodename);
1027         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1028         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1029                                      2, flashbase, 2, flashsize);
1030         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1031         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1032         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1033         g_free(nodename);
1034 
1035         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1036         qemu_fdt_add_subnode(vms->fdt, nodename);
1037         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1038         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1039                                      2, flashbase + flashsize, 2, flashsize);
1040         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1041         g_free(nodename);
1042     }
1043 }
1044 
1045 static bool virt_firmware_init(VirtMachineState *vms,
1046                                MemoryRegion *sysmem,
1047                                MemoryRegion *secure_sysmem)
1048 {
1049     int i;
1050     BlockBackend *pflash_blk0;
1051 
1052     /* Map legacy -drive if=pflash to machine properties */
1053     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1054         pflash_cfi01_legacy_drive(vms->flash[i],
1055                                   drive_get(IF_PFLASH, 0, i));
1056     }
1057 
1058     virt_flash_map(vms, sysmem, secure_sysmem);
1059 
1060     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1061 
1062     if (bios_name) {
1063         char *fname;
1064         MemoryRegion *mr;
1065         int image_size;
1066 
1067         if (pflash_blk0) {
1068             error_report("The contents of the first flash device may be "
1069                          "specified with -bios or with -drive if=pflash... "
1070                          "but you cannot use both options at once");
1071             exit(1);
1072         }
1073 
1074         /* Fall back to -bios */
1075 
1076         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1077         if (!fname) {
1078             error_report("Could not find ROM image '%s'", bios_name);
1079             exit(1);
1080         }
1081         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1082         image_size = load_image_mr(fname, mr);
1083         g_free(fname);
1084         if (image_size < 0) {
1085             error_report("Could not load ROM image '%s'", bios_name);
1086             exit(1);
1087         }
1088     }
1089 
1090     return pflash_blk0 || bios_name;
1091 }
1092 
1093 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1094 {
1095     MachineState *ms = MACHINE(vms);
1096     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1097     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1098     FWCfgState *fw_cfg;
1099     char *nodename;
1100 
1101     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1102     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1103 
1104     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1105     qemu_fdt_add_subnode(vms->fdt, nodename);
1106     qemu_fdt_setprop_string(vms->fdt, nodename,
1107                             "compatible", "qemu,fw-cfg-mmio");
1108     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1109                                  2, base, 2, size);
1110     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1111     g_free(nodename);
1112     return fw_cfg;
1113 }
1114 
1115 static void create_pcie_irq_map(const VirtMachineState *vms,
1116                                 uint32_t gic_phandle,
1117                                 int first_irq, const char *nodename)
1118 {
1119     int devfn, pin;
1120     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1121     uint32_t *irq_map = full_irq_map;
1122 
1123     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1124         for (pin = 0; pin < 4; pin++) {
1125             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1126             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1127             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1128             int i;
1129 
1130             uint32_t map[] = {
1131                 devfn << 8, 0, 0,                           /* devfn */
1132                 pin + 1,                                    /* PCI pin */
1133                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1134 
1135             /* Convert map to big endian */
1136             for (i = 0; i < 10; i++) {
1137                 irq_map[i] = cpu_to_be32(map[i]);
1138             }
1139             irq_map += 10;
1140         }
1141     }
1142 
1143     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
1144                      full_irq_map, sizeof(full_irq_map));
1145 
1146     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
1147                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
1148                            0x7           /* PCI irq */);
1149 }
1150 
1151 static void create_smmu(const VirtMachineState *vms,
1152                         PCIBus *bus)
1153 {
1154     char *node;
1155     const char compat[] = "arm,smmu-v3";
1156     int irq =  vms->irqmap[VIRT_SMMU];
1157     int i;
1158     hwaddr base = vms->memmap[VIRT_SMMU].base;
1159     hwaddr size = vms->memmap[VIRT_SMMU].size;
1160     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1161     DeviceState *dev;
1162 
1163     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1164         return;
1165     }
1166 
1167     dev = qdev_new("arm-smmuv3");
1168 
1169     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1170                              &error_abort);
1171     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1172     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1173     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1174         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1175                            qdev_get_gpio_in(vms->gic, irq + i));
1176     }
1177 
1178     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1179     qemu_fdt_add_subnode(vms->fdt, node);
1180     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1181     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
1182 
1183     qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
1184             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1185             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1186             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1187             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1188 
1189     qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
1190                      sizeof(irq_names));
1191 
1192     qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
1193     qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
1194     qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
1195 
1196     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1197 
1198     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1199     g_free(node);
1200 }
1201 
1202 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1203 {
1204     const char compat[] = "virtio,pci-iommu";
1205     uint16_t bdf = vms->virtio_iommu_bdf;
1206     char *node;
1207 
1208     vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1209 
1210     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1211     qemu_fdt_add_subnode(vms->fdt, node);
1212     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1213     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg",
1214                                  1, bdf << 8, 1, 0, 1, 0,
1215                                  1, 0, 1, 0);
1216 
1217     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1218     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1219     g_free(node);
1220 
1221     qemu_fdt_setprop_cells(vms->fdt, vms->pciehb_nodename, "iommu-map",
1222                            0x0, vms->iommu_phandle, 0x0, bdf,
1223                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1224 }
1225 
1226 static void create_pcie(VirtMachineState *vms)
1227 {
1228     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1229     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1230     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1231     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1232     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1233     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1234     hwaddr base_ecam, size_ecam;
1235     hwaddr base = base_mmio;
1236     int nr_pcie_buses;
1237     int irq = vms->irqmap[VIRT_PCIE];
1238     MemoryRegion *mmio_alias;
1239     MemoryRegion *mmio_reg;
1240     MemoryRegion *ecam_alias;
1241     MemoryRegion *ecam_reg;
1242     DeviceState *dev;
1243     char *nodename;
1244     int i, ecam_id;
1245     PCIHostState *pci;
1246 
1247     dev = qdev_new(TYPE_GPEX_HOST);
1248     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1249 
1250     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1251     base_ecam = vms->memmap[ecam_id].base;
1252     size_ecam = vms->memmap[ecam_id].size;
1253     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1254     /* Map only the first size_ecam bytes of ECAM space */
1255     ecam_alias = g_new0(MemoryRegion, 1);
1256     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1257     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1258                              ecam_reg, 0, size_ecam);
1259     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1260 
1261     /* Map the MMIO window into system address space so as to expose
1262      * the section of PCI MMIO space which starts at the same base address
1263      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1264      * the window).
1265      */
1266     mmio_alias = g_new0(MemoryRegion, 1);
1267     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1268     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1269                              mmio_reg, base_mmio, size_mmio);
1270     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1271 
1272     if (vms->highmem) {
1273         /* Map high MMIO space */
1274         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1275 
1276         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1277                                  mmio_reg, base_mmio_high, size_mmio_high);
1278         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1279                                     high_mmio_alias);
1280     }
1281 
1282     /* Map IO port space */
1283     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1284 
1285     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1286         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1287                            qdev_get_gpio_in(vms->gic, irq + i));
1288         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1289     }
1290 
1291     pci = PCI_HOST_BRIDGE(dev);
1292     if (pci->bus) {
1293         for (i = 0; i < nb_nics; i++) {
1294             NICInfo *nd = &nd_table[i];
1295 
1296             if (!nd->model) {
1297                 nd->model = g_strdup("virtio");
1298             }
1299 
1300             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1301         }
1302     }
1303 
1304     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1305     qemu_fdt_add_subnode(vms->fdt, nodename);
1306     qemu_fdt_setprop_string(vms->fdt, nodename,
1307                             "compatible", "pci-host-ecam-generic");
1308     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1309     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1310     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1311     qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1312     qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1313                            nr_pcie_buses - 1);
1314     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1315 
1316     if (vms->msi_phandle) {
1317         qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1318                                vms->msi_phandle);
1319     }
1320 
1321     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1322                                  2, base_ecam, 2, size_ecam);
1323 
1324     if (vms->highmem) {
1325         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1326                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1327                                      2, base_pio, 2, size_pio,
1328                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1329                                      2, base_mmio, 2, size_mmio,
1330                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1331                                      2, base_mmio_high,
1332                                      2, base_mmio_high, 2, size_mmio_high);
1333     } else {
1334         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1335                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1336                                      2, base_pio, 2, size_pio,
1337                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1338                                      2, base_mmio, 2, size_mmio);
1339     }
1340 
1341     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1342     create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1343 
1344     if (vms->iommu) {
1345         vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1346 
1347         switch (vms->iommu) {
1348         case VIRT_IOMMU_SMMUV3:
1349             create_smmu(vms, pci->bus);
1350             qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1351                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1352             break;
1353         default:
1354             g_assert_not_reached();
1355         }
1356     }
1357 }
1358 
1359 static void create_platform_bus(VirtMachineState *vms)
1360 {
1361     DeviceState *dev;
1362     SysBusDevice *s;
1363     int i;
1364     MemoryRegion *sysmem = get_system_memory();
1365 
1366     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1367     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1368     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1369     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1370     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1371     vms->platform_bus_dev = dev;
1372 
1373     s = SYS_BUS_DEVICE(dev);
1374     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1375         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1376         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1377     }
1378 
1379     memory_region_add_subregion(sysmem,
1380                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1381                                 sysbus_mmio_get_region(s, 0));
1382 }
1383 
1384 static void create_tag_ram(MemoryRegion *tag_sysmem,
1385                            hwaddr base, hwaddr size,
1386                            const char *name)
1387 {
1388     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1389 
1390     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1391     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1392 }
1393 
1394 static void create_secure_ram(VirtMachineState *vms,
1395                               MemoryRegion *secure_sysmem,
1396                               MemoryRegion *secure_tag_sysmem)
1397 {
1398     MemoryRegion *secram = g_new(MemoryRegion, 1);
1399     char *nodename;
1400     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1401     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1402 
1403     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1404                            &error_fatal);
1405     memory_region_add_subregion(secure_sysmem, base, secram);
1406 
1407     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1408     qemu_fdt_add_subnode(vms->fdt, nodename);
1409     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1410     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1411     qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1412     qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1413 
1414     if (secure_tag_sysmem) {
1415         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1416     }
1417 
1418     g_free(nodename);
1419 }
1420 
1421 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1422 {
1423     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1424                                                  bootinfo);
1425 
1426     *fdt_size = board->fdt_size;
1427     return board->fdt;
1428 }
1429 
1430 static void virt_build_smbios(VirtMachineState *vms)
1431 {
1432     MachineClass *mc = MACHINE_GET_CLASS(vms);
1433     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1434     uint8_t *smbios_tables, *smbios_anchor;
1435     size_t smbios_tables_len, smbios_anchor_len;
1436     const char *product = "QEMU Virtual Machine";
1437 
1438     if (kvm_enabled()) {
1439         product = "KVM Virtual Machine";
1440     }
1441 
1442     smbios_set_defaults("QEMU", product,
1443                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1444                         true, SMBIOS_ENTRY_POINT_30);
1445 
1446     smbios_get_tables(MACHINE(vms), NULL, 0, &smbios_tables, &smbios_tables_len,
1447                       &smbios_anchor, &smbios_anchor_len);
1448 
1449     if (smbios_anchor) {
1450         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1451                         smbios_tables, smbios_tables_len);
1452         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1453                         smbios_anchor, smbios_anchor_len);
1454     }
1455 }
1456 
1457 static
1458 void virt_machine_done(Notifier *notifier, void *data)
1459 {
1460     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1461                                          machine_done);
1462     MachineState *ms = MACHINE(vms);
1463     ARMCPU *cpu = ARM_CPU(first_cpu);
1464     struct arm_boot_info *info = &vms->bootinfo;
1465     AddressSpace *as = arm_boot_address_space(cpu, info);
1466 
1467     /*
1468      * If the user provided a dtb, we assume the dynamic sysbus nodes
1469      * already are integrated there. This corresponds to a use case where
1470      * the dynamic sysbus nodes are complex and their generation is not yet
1471      * supported. In that case the user can take charge of the guest dt
1472      * while qemu takes charge of the qom stuff.
1473      */
1474     if (info->dtb_filename == NULL) {
1475         platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1476                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1477                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1478                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1479     }
1480     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1481         exit(1);
1482     }
1483 
1484     virt_acpi_setup(vms);
1485     virt_build_smbios(vms);
1486 }
1487 
1488 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1489 {
1490     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1491     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1492 
1493     if (!vmc->disallow_affinity_adjustment) {
1494         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1495          * GIC's target-list limitations. 32-bit KVM hosts currently
1496          * always create clusters of 4 CPUs, but that is expected to
1497          * change when they gain support for gicv3. When KVM is enabled
1498          * it will override the changes we make here, therefore our
1499          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1500          * and to improve SGI efficiency.
1501          */
1502         if (vms->gic_version == VIRT_GIC_VERSION_3) {
1503             clustersz = GICV3_TARGETLIST_BITS;
1504         } else {
1505             clustersz = GIC_TARGETLIST_BITS;
1506         }
1507     }
1508     return arm_cpu_mp_affinity(idx, clustersz);
1509 }
1510 
1511 static void virt_set_memmap(VirtMachineState *vms)
1512 {
1513     MachineState *ms = MACHINE(vms);
1514     hwaddr base, device_memory_base, device_memory_size;
1515     int i;
1516 
1517     vms->memmap = extended_memmap;
1518 
1519     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1520         vms->memmap[i] = base_memmap[i];
1521     }
1522 
1523     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1524         error_report("unsupported number of memory slots: %"PRIu64,
1525                      ms->ram_slots);
1526         exit(EXIT_FAILURE);
1527     }
1528 
1529     /*
1530      * We compute the base of the high IO region depending on the
1531      * amount of initial and device memory. The device memory start/size
1532      * is aligned on 1GiB. We never put the high IO region below 256GiB
1533      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1534      * The device region size assumes 1GiB page max alignment per slot.
1535      */
1536     device_memory_base =
1537         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1538     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1539 
1540     /* Base address of the high IO region */
1541     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1542     if (base < device_memory_base) {
1543         error_report("maxmem/slots too huge");
1544         exit(EXIT_FAILURE);
1545     }
1546     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1547         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1548     }
1549 
1550     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1551         hwaddr size = extended_memmap[i].size;
1552 
1553         base = ROUND_UP(base, size);
1554         vms->memmap[i].base = base;
1555         vms->memmap[i].size = size;
1556         base += size;
1557     }
1558     vms->highest_gpa = base - 1;
1559     if (device_memory_size > 0) {
1560         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1561         ms->device_memory->base = device_memory_base;
1562         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1563                            "device-memory", device_memory_size);
1564     }
1565 }
1566 
1567 /*
1568  * finalize_gic_version - Determines the final gic_version
1569  * according to the gic-version property
1570  *
1571  * Default GIC type is v2
1572  */
1573 static void finalize_gic_version(VirtMachineState *vms)
1574 {
1575     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1576 
1577     if (kvm_enabled()) {
1578         int probe_bitmap;
1579 
1580         if (!kvm_irqchip_in_kernel()) {
1581             switch (vms->gic_version) {
1582             case VIRT_GIC_VERSION_HOST:
1583                 warn_report(
1584                     "gic-version=host not relevant with kernel-irqchip=off "
1585                      "as only userspace GICv2 is supported. Using v2 ...");
1586                 return;
1587             case VIRT_GIC_VERSION_MAX:
1588             case VIRT_GIC_VERSION_NOSEL:
1589                 vms->gic_version = VIRT_GIC_VERSION_2;
1590                 return;
1591             case VIRT_GIC_VERSION_2:
1592                 return;
1593             case VIRT_GIC_VERSION_3:
1594                 error_report(
1595                     "gic-version=3 is not supported with kernel-irqchip=off");
1596                 exit(1);
1597             }
1598         }
1599 
1600         probe_bitmap = kvm_arm_vgic_probe();
1601         if (!probe_bitmap) {
1602             error_report("Unable to determine GIC version supported by host");
1603             exit(1);
1604         }
1605 
1606         switch (vms->gic_version) {
1607         case VIRT_GIC_VERSION_HOST:
1608         case VIRT_GIC_VERSION_MAX:
1609             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1610                 vms->gic_version = VIRT_GIC_VERSION_3;
1611             } else {
1612                 vms->gic_version = VIRT_GIC_VERSION_2;
1613             }
1614             return;
1615         case VIRT_GIC_VERSION_NOSEL:
1616             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1617                 vms->gic_version = VIRT_GIC_VERSION_2;
1618             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1619                 /*
1620                  * in case the host does not support v2 in-kernel emulation or
1621                  * the end-user requested more than 8 VCPUs we now default
1622                  * to v3. In any case defaulting to v2 would be broken.
1623                  */
1624                 vms->gic_version = VIRT_GIC_VERSION_3;
1625             } else if (max_cpus > GIC_NCPU) {
1626                 error_report("host only supports in-kernel GICv2 emulation "
1627                              "but more than 8 vcpus are requested");
1628                 exit(1);
1629             }
1630             break;
1631         case VIRT_GIC_VERSION_2:
1632         case VIRT_GIC_VERSION_3:
1633             break;
1634         }
1635 
1636         /* Check chosen version is effectively supported by the host */
1637         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1638             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1639             error_report("host does not support in-kernel GICv2 emulation");
1640             exit(1);
1641         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1642                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1643             error_report("host does not support in-kernel GICv3 emulation");
1644             exit(1);
1645         }
1646         return;
1647     }
1648 
1649     /* TCG mode */
1650     switch (vms->gic_version) {
1651     case VIRT_GIC_VERSION_NOSEL:
1652         vms->gic_version = VIRT_GIC_VERSION_2;
1653         break;
1654     case VIRT_GIC_VERSION_MAX:
1655         vms->gic_version = VIRT_GIC_VERSION_3;
1656         break;
1657     case VIRT_GIC_VERSION_HOST:
1658         error_report("gic-version=host requires KVM");
1659         exit(1);
1660     case VIRT_GIC_VERSION_2:
1661     case VIRT_GIC_VERSION_3:
1662         break;
1663     }
1664 }
1665 
1666 /*
1667  * virt_cpu_post_init() must be called after the CPUs have
1668  * been realized and the GIC has been created.
1669  */
1670 static void virt_cpu_post_init(VirtMachineState *vms, int max_cpus,
1671                                MemoryRegion *sysmem)
1672 {
1673     bool aarch64, pmu, steal_time;
1674     CPUState *cpu;
1675 
1676     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1677     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1678     steal_time = object_property_get_bool(OBJECT(first_cpu),
1679                                           "kvm-steal-time", NULL);
1680 
1681     if (kvm_enabled()) {
1682         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1683         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1684 
1685         if (steal_time) {
1686             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1687             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1688 
1689             /* The memory region size must be a multiple of host page size. */
1690             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1691 
1692             if (pvtime_size > pvtime_reg_size) {
1693                 error_report("pvtime requires a %" HWADDR_PRId
1694                              " byte memory region for %d CPUs,"
1695                              " but only %" HWADDR_PRId " has been reserved",
1696                              pvtime_size, max_cpus, pvtime_reg_size);
1697                 exit(1);
1698             }
1699 
1700             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1701             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1702         }
1703 
1704         CPU_FOREACH(cpu) {
1705             if (pmu) {
1706                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1707                 if (kvm_irqchip_in_kernel()) {
1708                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1709                 }
1710                 kvm_arm_pmu_init(cpu);
1711             }
1712             if (steal_time) {
1713                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1714                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1715             }
1716         }
1717     } else {
1718         if (aarch64 && vms->highmem) {
1719             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1720             int pamax = arm_pamax(ARM_CPU(first_cpu));
1721 
1722             if (pamax < requested_pa_size) {
1723                 error_report("VCPU supports less PA bits (%d) than "
1724                              "requested by the memory map (%d)",
1725                              pamax, requested_pa_size);
1726                 exit(1);
1727             }
1728         }
1729     }
1730 }
1731 
1732 static void machvirt_init(MachineState *machine)
1733 {
1734     VirtMachineState *vms = VIRT_MACHINE(machine);
1735     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1736     MachineClass *mc = MACHINE_GET_CLASS(machine);
1737     const CPUArchIdList *possible_cpus;
1738     MemoryRegion *sysmem = get_system_memory();
1739     MemoryRegion *secure_sysmem = NULL;
1740     MemoryRegion *tag_sysmem = NULL;
1741     MemoryRegion *secure_tag_sysmem = NULL;
1742     int n, virt_max_cpus;
1743     bool firmware_loaded;
1744     bool aarch64 = true;
1745     bool has_ged = !vmc->no_ged;
1746     unsigned int smp_cpus = machine->smp.cpus;
1747     unsigned int max_cpus = machine->smp.max_cpus;
1748 
1749     /*
1750      * In accelerated mode, the memory map is computed earlier in kvm_type()
1751      * to create a VM with the right number of IPA bits.
1752      */
1753     if (!vms->memmap) {
1754         virt_set_memmap(vms);
1755     }
1756 
1757     /* We can probe only here because during property set
1758      * KVM is not available yet
1759      */
1760     finalize_gic_version(vms);
1761 
1762     if (!cpu_type_valid(machine->cpu_type)) {
1763         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1764         exit(1);
1765     }
1766 
1767     if (vms->secure) {
1768         if (kvm_enabled()) {
1769             error_report("mach-virt: KVM does not support Security extensions");
1770             exit(1);
1771         }
1772 
1773         /*
1774          * The Secure view of the world is the same as the NonSecure,
1775          * but with a few extra devices. Create it as a container region
1776          * containing the system memory at low priority; any secure-only
1777          * devices go in at higher priority and take precedence.
1778          */
1779         secure_sysmem = g_new(MemoryRegion, 1);
1780         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1781                            UINT64_MAX);
1782         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1783     }
1784 
1785     firmware_loaded = virt_firmware_init(vms, sysmem,
1786                                          secure_sysmem ?: sysmem);
1787 
1788     /* If we have an EL3 boot ROM then the assumption is that it will
1789      * implement PSCI itself, so disable QEMU's internal implementation
1790      * so it doesn't get in the way. Instead of starting secondary
1791      * CPUs in PSCI powerdown state we will start them all running and
1792      * let the boot ROM sort them out.
1793      * The usual case is that we do use QEMU's PSCI implementation;
1794      * if the guest has EL2 then we will use SMC as the conduit,
1795      * and otherwise we will use HVC (for backwards compatibility and
1796      * because if we're using KVM then we must use HVC).
1797      */
1798     if (vms->secure && firmware_loaded) {
1799         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1800     } else if (vms->virt) {
1801         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1802     } else {
1803         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1804     }
1805 
1806     /* The maximum number of CPUs depends on the GIC version, or on how
1807      * many redistributors we can fit into the memory map.
1808      */
1809     if (vms->gic_version == VIRT_GIC_VERSION_3) {
1810         virt_max_cpus =
1811             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1812         virt_max_cpus +=
1813             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1814     } else {
1815         virt_max_cpus = GIC_NCPU;
1816     }
1817 
1818     if (max_cpus > virt_max_cpus) {
1819         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1820                      "supported by machine 'mach-virt' (%d)",
1821                      max_cpus, virt_max_cpus);
1822         exit(1);
1823     }
1824 
1825     vms->smp_cpus = smp_cpus;
1826 
1827     if (vms->virt && kvm_enabled()) {
1828         error_report("mach-virt: KVM does not support providing "
1829                      "Virtualization extensions to the guest CPU");
1830         exit(1);
1831     }
1832 
1833     if (vms->mte && kvm_enabled()) {
1834         error_report("mach-virt: KVM does not support providing "
1835                      "MTE to the guest CPU");
1836         exit(1);
1837     }
1838 
1839     create_fdt(vms);
1840 
1841     possible_cpus = mc->possible_cpu_arch_ids(machine);
1842     for (n = 0; n < possible_cpus->len; n++) {
1843         Object *cpuobj;
1844         CPUState *cs;
1845 
1846         if (n >= smp_cpus) {
1847             break;
1848         }
1849 
1850         cpuobj = object_new(possible_cpus->cpus[n].type);
1851         object_property_set_int(cpuobj, "mp-affinity",
1852                                 possible_cpus->cpus[n].arch_id, NULL);
1853 
1854         cs = CPU(cpuobj);
1855         cs->cpu_index = n;
1856 
1857         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1858                           &error_fatal);
1859 
1860         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1861 
1862         if (!vms->secure) {
1863             object_property_set_bool(cpuobj, "has_el3", false, NULL);
1864         }
1865 
1866         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
1867             object_property_set_bool(cpuobj, "has_el2", false, NULL);
1868         }
1869 
1870         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1871             object_property_set_int(cpuobj, "psci-conduit", vms->psci_conduit,
1872                                     NULL);
1873 
1874             /* Secondary CPUs start in PSCI powered-down state */
1875             if (n > 0) {
1876                 object_property_set_bool(cpuobj, "start-powered-off", true,
1877                                          NULL);
1878             }
1879         }
1880 
1881         if (vmc->kvm_no_adjvtime &&
1882             object_property_find(cpuobj, "kvm-no-adjvtime")) {
1883             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
1884         }
1885 
1886         if (vmc->no_kvm_steal_time &&
1887             object_property_find(cpuobj, "kvm-steal-time")) {
1888             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
1889         }
1890 
1891         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
1892             object_property_set_bool(cpuobj, "pmu", false, NULL);
1893         }
1894 
1895         if (object_property_find(cpuobj, "reset-cbar")) {
1896             object_property_set_int(cpuobj, "reset-cbar",
1897                                     vms->memmap[VIRT_CPUPERIPHS].base,
1898                                     &error_abort);
1899         }
1900 
1901         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
1902                                  &error_abort);
1903         if (vms->secure) {
1904             object_property_set_link(cpuobj, "secure-memory",
1905                                      OBJECT(secure_sysmem), &error_abort);
1906         }
1907 
1908         if (vms->mte) {
1909             /* Create the memory region only once, but link to all cpus. */
1910             if (!tag_sysmem) {
1911                 /*
1912                  * The property exists only if MemTag is supported.
1913                  * If it is, we must allocate the ram to back that up.
1914                  */
1915                 if (!object_property_find(cpuobj, "tag-memory")) {
1916                     error_report("MTE requested, but not supported "
1917                                  "by the guest CPU");
1918                     exit(1);
1919                 }
1920 
1921                 tag_sysmem = g_new(MemoryRegion, 1);
1922                 memory_region_init(tag_sysmem, OBJECT(machine),
1923                                    "tag-memory", UINT64_MAX / 32);
1924 
1925                 if (vms->secure) {
1926                     secure_tag_sysmem = g_new(MemoryRegion, 1);
1927                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
1928                                        "secure-tag-memory", UINT64_MAX / 32);
1929 
1930                     /* As with ram, secure-tag takes precedence over tag.  */
1931                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
1932                                                         tag_sysmem, -1);
1933                 }
1934             }
1935 
1936             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
1937                                      &error_abort);
1938             if (vms->secure) {
1939                 object_property_set_link(cpuobj, "secure-tag-memory",
1940                                          OBJECT(secure_tag_sysmem),
1941                                          &error_abort);
1942             }
1943         }
1944 
1945         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
1946         object_unref(cpuobj);
1947     }
1948     fdt_add_timer_nodes(vms);
1949     fdt_add_cpu_nodes(vms);
1950 
1951     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
1952                                 machine->ram);
1953     if (machine->device_memory) {
1954         memory_region_add_subregion(sysmem, machine->device_memory->base,
1955                                     &machine->device_memory->mr);
1956     }
1957 
1958     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
1959 
1960     create_gic(vms);
1961 
1962     virt_cpu_post_init(vms, possible_cpus->len, sysmem);
1963 
1964     fdt_add_pmu_nodes(vms);
1965 
1966     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
1967 
1968     if (vms->secure) {
1969         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
1970         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
1971     }
1972 
1973     if (tag_sysmem) {
1974         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
1975                        machine->ram_size, "mach-virt.tag");
1976     }
1977 
1978     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
1979 
1980     create_rtc(vms);
1981 
1982     create_pcie(vms);
1983 
1984     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
1985         vms->acpi_dev = create_acpi_ged(vms);
1986     } else {
1987         create_gpio(vms);
1988     }
1989 
1990      /* connect powerdown request */
1991      vms->powerdown_notifier.notify = virt_powerdown_req;
1992      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
1993 
1994     /* Create mmio transports, so the user can create virtio backends
1995      * (which will be automatically plugged in to the transports). If
1996      * no backend is created the transport will just sit harmlessly idle.
1997      */
1998     create_virtio_devices(vms);
1999 
2000     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2001     rom_set_fw(vms->fw_cfg);
2002 
2003     create_platform_bus(vms);
2004 
2005     if (machine->nvdimms_state->is_enabled) {
2006         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2007             .space_id = AML_AS_SYSTEM_MEMORY,
2008             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2009             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2010         };
2011 
2012         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2013                                arm_virt_nvdimm_acpi_dsmio,
2014                                vms->fw_cfg, OBJECT(vms));
2015     }
2016 
2017     vms->bootinfo.ram_size = machine->ram_size;
2018     vms->bootinfo.nb_cpus = smp_cpus;
2019     vms->bootinfo.board_id = -1;
2020     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2021     vms->bootinfo.get_dtb = machvirt_dtb;
2022     vms->bootinfo.skip_dtb_autoload = true;
2023     vms->bootinfo.firmware_loaded = firmware_loaded;
2024     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2025 
2026     vms->machine_done.notify = virt_machine_done;
2027     qemu_add_machine_init_done_notifier(&vms->machine_done);
2028 }
2029 
2030 static bool virt_get_secure(Object *obj, Error **errp)
2031 {
2032     VirtMachineState *vms = VIRT_MACHINE(obj);
2033 
2034     return vms->secure;
2035 }
2036 
2037 static void virt_set_secure(Object *obj, bool value, Error **errp)
2038 {
2039     VirtMachineState *vms = VIRT_MACHINE(obj);
2040 
2041     vms->secure = value;
2042 }
2043 
2044 static bool virt_get_virt(Object *obj, Error **errp)
2045 {
2046     VirtMachineState *vms = VIRT_MACHINE(obj);
2047 
2048     return vms->virt;
2049 }
2050 
2051 static void virt_set_virt(Object *obj, bool value, Error **errp)
2052 {
2053     VirtMachineState *vms = VIRT_MACHINE(obj);
2054 
2055     vms->virt = value;
2056 }
2057 
2058 static bool virt_get_highmem(Object *obj, Error **errp)
2059 {
2060     VirtMachineState *vms = VIRT_MACHINE(obj);
2061 
2062     return vms->highmem;
2063 }
2064 
2065 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2066 {
2067     VirtMachineState *vms = VIRT_MACHINE(obj);
2068 
2069     vms->highmem = value;
2070 }
2071 
2072 static bool virt_get_its(Object *obj, Error **errp)
2073 {
2074     VirtMachineState *vms = VIRT_MACHINE(obj);
2075 
2076     return vms->its;
2077 }
2078 
2079 static void virt_set_its(Object *obj, bool value, Error **errp)
2080 {
2081     VirtMachineState *vms = VIRT_MACHINE(obj);
2082 
2083     vms->its = value;
2084 }
2085 
2086 bool virt_is_acpi_enabled(VirtMachineState *vms)
2087 {
2088     if (vms->acpi == ON_OFF_AUTO_OFF) {
2089         return false;
2090     }
2091     return true;
2092 }
2093 
2094 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2095                           void *opaque, Error **errp)
2096 {
2097     VirtMachineState *vms = VIRT_MACHINE(obj);
2098     OnOffAuto acpi = vms->acpi;
2099 
2100     visit_type_OnOffAuto(v, name, &acpi, errp);
2101 }
2102 
2103 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2104                           void *opaque, Error **errp)
2105 {
2106     VirtMachineState *vms = VIRT_MACHINE(obj);
2107 
2108     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2109 }
2110 
2111 static bool virt_get_ras(Object *obj, Error **errp)
2112 {
2113     VirtMachineState *vms = VIRT_MACHINE(obj);
2114 
2115     return vms->ras;
2116 }
2117 
2118 static void virt_set_ras(Object *obj, bool value, Error **errp)
2119 {
2120     VirtMachineState *vms = VIRT_MACHINE(obj);
2121 
2122     vms->ras = value;
2123 }
2124 
2125 static bool virt_get_mte(Object *obj, Error **errp)
2126 {
2127     VirtMachineState *vms = VIRT_MACHINE(obj);
2128 
2129     return vms->mte;
2130 }
2131 
2132 static void virt_set_mte(Object *obj, bool value, Error **errp)
2133 {
2134     VirtMachineState *vms = VIRT_MACHINE(obj);
2135 
2136     vms->mte = value;
2137 }
2138 
2139 static char *virt_get_gic_version(Object *obj, Error **errp)
2140 {
2141     VirtMachineState *vms = VIRT_MACHINE(obj);
2142     const char *val = vms->gic_version == VIRT_GIC_VERSION_3 ? "3" : "2";
2143 
2144     return g_strdup(val);
2145 }
2146 
2147 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2148 {
2149     VirtMachineState *vms = VIRT_MACHINE(obj);
2150 
2151     if (!strcmp(value, "3")) {
2152         vms->gic_version = VIRT_GIC_VERSION_3;
2153     } else if (!strcmp(value, "2")) {
2154         vms->gic_version = VIRT_GIC_VERSION_2;
2155     } else if (!strcmp(value, "host")) {
2156         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2157     } else if (!strcmp(value, "max")) {
2158         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2159     } else {
2160         error_setg(errp, "Invalid gic-version value");
2161         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2162     }
2163 }
2164 
2165 static char *virt_get_iommu(Object *obj, Error **errp)
2166 {
2167     VirtMachineState *vms = VIRT_MACHINE(obj);
2168 
2169     switch (vms->iommu) {
2170     case VIRT_IOMMU_NONE:
2171         return g_strdup("none");
2172     case VIRT_IOMMU_SMMUV3:
2173         return g_strdup("smmuv3");
2174     default:
2175         g_assert_not_reached();
2176     }
2177 }
2178 
2179 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2180 {
2181     VirtMachineState *vms = VIRT_MACHINE(obj);
2182 
2183     if (!strcmp(value, "smmuv3")) {
2184         vms->iommu = VIRT_IOMMU_SMMUV3;
2185     } else if (!strcmp(value, "none")) {
2186         vms->iommu = VIRT_IOMMU_NONE;
2187     } else {
2188         error_setg(errp, "Invalid iommu value");
2189         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2190     }
2191 }
2192 
2193 static CpuInstanceProperties
2194 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2195 {
2196     MachineClass *mc = MACHINE_GET_CLASS(ms);
2197     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2198 
2199     assert(cpu_index < possible_cpus->len);
2200     return possible_cpus->cpus[cpu_index].props;
2201 }
2202 
2203 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2204 {
2205     return idx % ms->numa_state->num_nodes;
2206 }
2207 
2208 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2209 {
2210     int n;
2211     unsigned int max_cpus = ms->smp.max_cpus;
2212     VirtMachineState *vms = VIRT_MACHINE(ms);
2213 
2214     if (ms->possible_cpus) {
2215         assert(ms->possible_cpus->len == max_cpus);
2216         return ms->possible_cpus;
2217     }
2218 
2219     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2220                                   sizeof(CPUArchId) * max_cpus);
2221     ms->possible_cpus->len = max_cpus;
2222     for (n = 0; n < ms->possible_cpus->len; n++) {
2223         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2224         ms->possible_cpus->cpus[n].arch_id =
2225             virt_cpu_mp_affinity(vms, n);
2226         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2227         ms->possible_cpus->cpus[n].props.thread_id = n;
2228     }
2229     return ms->possible_cpus;
2230 }
2231 
2232 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2233                                  Error **errp)
2234 {
2235     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2236     const MachineState *ms = MACHINE(hotplug_dev);
2237     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2238 
2239     if (!vms->acpi_dev) {
2240         error_setg(errp,
2241                    "memory hotplug is not enabled: missing acpi-ged device");
2242         return;
2243     }
2244 
2245     if (vms->mte) {
2246         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2247         return;
2248     }
2249 
2250     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2251         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2252         return;
2253     }
2254 
2255     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2256 }
2257 
2258 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2259                              DeviceState *dev, Error **errp)
2260 {
2261     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2262     MachineState *ms = MACHINE(hotplug_dev);
2263     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2264 
2265     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2266 
2267     if (is_nvdimm) {
2268         nvdimm_plug(ms->nvdimms_state);
2269     }
2270 
2271     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2272                          dev, &error_abort);
2273 }
2274 
2275 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2276                                             DeviceState *dev, Error **errp)
2277 {
2278     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2279 
2280     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2281         virt_memory_pre_plug(hotplug_dev, dev, errp);
2282     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2283         hwaddr db_start = 0, db_end = 0;
2284         char *resv_prop_str;
2285 
2286         switch (vms->msi_controller) {
2287         case VIRT_MSI_CTRL_NONE:
2288             return;
2289         case VIRT_MSI_CTRL_ITS:
2290             /* GITS_TRANSLATER page */
2291             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2292             db_end = base_memmap[VIRT_GIC_ITS].base +
2293                      base_memmap[VIRT_GIC_ITS].size - 1;
2294             break;
2295         case VIRT_MSI_CTRL_GICV2M:
2296             /* MSI_SETSPI_NS page */
2297             db_start = base_memmap[VIRT_GIC_V2M].base;
2298             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2299             break;
2300         }
2301         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2302                                         db_start, db_end,
2303                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2304 
2305         qdev_prop_set_uint32(dev, "len-reserved-regions", 1);
2306         qdev_prop_set_string(dev, "reserved-regions[0]", resv_prop_str);
2307         g_free(resv_prop_str);
2308     }
2309 }
2310 
2311 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2312                                         DeviceState *dev, Error **errp)
2313 {
2314     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2315 
2316     if (vms->platform_bus_dev) {
2317         if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
2318             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2319                                      SYS_BUS_DEVICE(dev));
2320         }
2321     }
2322     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2323         virt_memory_plug(hotplug_dev, dev, errp);
2324     }
2325     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2326         PCIDevice *pdev = PCI_DEVICE(dev);
2327 
2328         vms->iommu = VIRT_IOMMU_VIRTIO;
2329         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2330         create_virtio_iommu_dt_bindings(vms);
2331     }
2332 }
2333 
2334 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2335                                      DeviceState *dev, Error **errp)
2336 {
2337     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2338     Error *local_err = NULL;
2339 
2340     if (!vms->acpi_dev) {
2341         error_setg(&local_err,
2342                    "memory hotplug is not enabled: missing acpi-ged device");
2343         goto out;
2344     }
2345 
2346     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2347         error_setg(&local_err,
2348                    "nvdimm device hot unplug is not supported yet.");
2349         goto out;
2350     }
2351 
2352     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2353                                    &local_err);
2354 out:
2355     error_propagate(errp, local_err);
2356 }
2357 
2358 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2359                              DeviceState *dev, Error **errp)
2360 {
2361     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2362     Error *local_err = NULL;
2363 
2364     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2365     if (local_err) {
2366         goto out;
2367     }
2368 
2369     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2370     qdev_unrealize(dev);
2371 
2372 out:
2373     error_propagate(errp, local_err);
2374 }
2375 
2376 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2377                                           DeviceState *dev, Error **errp)
2378 {
2379     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2380         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2381     } else {
2382         error_setg(errp, "device unplug request for unsupported device"
2383                    " type: %s", object_get_typename(OBJECT(dev)));
2384     }
2385 }
2386 
2387 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2388                                           DeviceState *dev, Error **errp)
2389 {
2390     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2391         virt_dimm_unplug(hotplug_dev, dev, errp);
2392     } else {
2393         error_setg(errp, "virt: device unplug for unsupported device"
2394                    " type: %s", object_get_typename(OBJECT(dev)));
2395     }
2396 }
2397 
2398 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2399                                                         DeviceState *dev)
2400 {
2401     if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE) ||
2402        (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
2403         return HOTPLUG_HANDLER(machine);
2404     }
2405     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2406         VirtMachineState *vms = VIRT_MACHINE(machine);
2407 
2408         if (!vms->bootinfo.firmware_loaded || !virt_is_acpi_enabled(vms)) {
2409             return HOTPLUG_HANDLER(machine);
2410         }
2411     }
2412     return NULL;
2413 }
2414 
2415 /*
2416  * for arm64 kvm_type [7-0] encodes the requested number of bits
2417  * in the IPA address space
2418  */
2419 static int virt_kvm_type(MachineState *ms, const char *type_str)
2420 {
2421     VirtMachineState *vms = VIRT_MACHINE(ms);
2422     int max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms);
2423     int requested_pa_size;
2424 
2425     /* we freeze the memory map to compute the highest gpa */
2426     virt_set_memmap(vms);
2427 
2428     requested_pa_size = 64 - clz64(vms->highest_gpa);
2429 
2430     if (requested_pa_size > max_vm_pa_size) {
2431         error_report("-m and ,maxmem option values "
2432                      "require an IPA range (%d bits) larger than "
2433                      "the one supported by the host (%d bits)",
2434                      requested_pa_size, max_vm_pa_size);
2435        exit(1);
2436     }
2437     /*
2438      * By default we return 0 which corresponds to an implicit legacy
2439      * 40b IPA setting. Otherwise we return the actual requested PA
2440      * logsize
2441      */
2442     return requested_pa_size > 40 ? requested_pa_size : 0;
2443 }
2444 
2445 static void virt_machine_class_init(ObjectClass *oc, void *data)
2446 {
2447     MachineClass *mc = MACHINE_CLASS(oc);
2448     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2449 
2450     mc->init = machvirt_init;
2451     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2452      * The value may be reduced later when we have more information about the
2453      * configuration of the particular instance.
2454      */
2455     mc->max_cpus = 512;
2456     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2457     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2458     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2459     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2460     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2461     mc->block_default_type = IF_VIRTIO;
2462     mc->no_cdrom = 1;
2463     mc->pci_allow_0_address = true;
2464     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2465     mc->minimum_page_bits = 12;
2466     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2467     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2468     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2469     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2470     mc->kvm_type = virt_kvm_type;
2471     assert(!mc->get_hotplug_handler);
2472     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2473     hc->pre_plug = virt_machine_device_pre_plug_cb;
2474     hc->plug = virt_machine_device_plug_cb;
2475     hc->unplug_request = virt_machine_device_unplug_request_cb;
2476     hc->unplug = virt_machine_device_unplug_cb;
2477     mc->nvdimm_supported = true;
2478     mc->auto_enable_numa_with_memhp = true;
2479     mc->auto_enable_numa_with_memdev = true;
2480     mc->default_ram_id = "mach-virt.ram";
2481 
2482     object_class_property_add(oc, "acpi", "OnOffAuto",
2483         virt_get_acpi, virt_set_acpi,
2484         NULL, NULL);
2485     object_class_property_set_description(oc, "acpi",
2486         "Enable ACPI");
2487 }
2488 
2489 static void virt_instance_init(Object *obj)
2490 {
2491     VirtMachineState *vms = VIRT_MACHINE(obj);
2492     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2493 
2494     /* EL3 is disabled by default on virt: this makes us consistent
2495      * between KVM and TCG for this board, and it also allows us to
2496      * boot UEFI blobs which assume no TrustZone support.
2497      */
2498     vms->secure = false;
2499     object_property_add_bool(obj, "secure", virt_get_secure,
2500                              virt_set_secure);
2501     object_property_set_description(obj, "secure",
2502                                     "Set on/off to enable/disable the ARM "
2503                                     "Security Extensions (TrustZone)");
2504 
2505     /* EL2 is also disabled by default, for similar reasons */
2506     vms->virt = false;
2507     object_property_add_bool(obj, "virtualization", virt_get_virt,
2508                              virt_set_virt);
2509     object_property_set_description(obj, "virtualization",
2510                                     "Set on/off to enable/disable emulating a "
2511                                     "guest CPU which implements the ARM "
2512                                     "Virtualization Extensions");
2513 
2514     /* High memory is enabled by default */
2515     vms->highmem = true;
2516     object_property_add_bool(obj, "highmem", virt_get_highmem,
2517                              virt_set_highmem);
2518     object_property_set_description(obj, "highmem",
2519                                     "Set on/off to enable/disable using "
2520                                     "physical address space above 32 bits");
2521     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
2522     object_property_add_str(obj, "gic-version", virt_get_gic_version,
2523                         virt_set_gic_version);
2524     object_property_set_description(obj, "gic-version",
2525                                     "Set GIC version. "
2526                                     "Valid values are 2, 3, host and max");
2527 
2528     vms->highmem_ecam = !vmc->no_highmem_ecam;
2529 
2530     if (vmc->no_its) {
2531         vms->its = false;
2532     } else {
2533         /* Default allows ITS instantiation */
2534         vms->its = true;
2535         object_property_add_bool(obj, "its", virt_get_its,
2536                                  virt_set_its);
2537         object_property_set_description(obj, "its",
2538                                         "Set on/off to enable/disable "
2539                                         "ITS instantiation");
2540     }
2541 
2542     /* Default disallows iommu instantiation */
2543     vms->iommu = VIRT_IOMMU_NONE;
2544     object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu);
2545     object_property_set_description(obj, "iommu",
2546                                     "Set the IOMMU type. "
2547                                     "Valid values are none and smmuv3");
2548 
2549     /* Default disallows RAS instantiation */
2550     vms->ras = false;
2551     object_property_add_bool(obj, "ras", virt_get_ras,
2552                              virt_set_ras);
2553     object_property_set_description(obj, "ras",
2554                                     "Set on/off to enable/disable reporting host memory errors "
2555                                     "to a KVM guest using ACPI and guest external abort exceptions");
2556 
2557     /* MTE is disabled by default.  */
2558     vms->mte = false;
2559     object_property_add_bool(obj, "mte", virt_get_mte, virt_set_mte);
2560     object_property_set_description(obj, "mte",
2561                                     "Set on/off to enable/disable emulating a "
2562                                     "guest CPU which implements the ARM "
2563                                     "Memory Tagging Extension");
2564 
2565     vms->irqmap = a15irqmap;
2566 
2567     virt_flash_create(vms);
2568 }
2569 
2570 static const TypeInfo virt_machine_info = {
2571     .name          = TYPE_VIRT_MACHINE,
2572     .parent        = TYPE_MACHINE,
2573     .abstract      = true,
2574     .instance_size = sizeof(VirtMachineState),
2575     .class_size    = sizeof(VirtMachineClass),
2576     .class_init    = virt_machine_class_init,
2577     .instance_init = virt_instance_init,
2578     .interfaces = (InterfaceInfo[]) {
2579          { TYPE_HOTPLUG_HANDLER },
2580          { }
2581     },
2582 };
2583 
2584 static void machvirt_machine_init(void)
2585 {
2586     type_register_static(&virt_machine_info);
2587 }
2588 type_init(machvirt_machine_init);
2589 
2590 static void virt_machine_5_2_options(MachineClass *mc)
2591 {
2592 }
2593 DEFINE_VIRT_MACHINE_AS_LATEST(5, 2)
2594 
2595 static void virt_machine_5_1_options(MachineClass *mc)
2596 {
2597     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2598 
2599     virt_machine_5_2_options(mc);
2600     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
2601     vmc->no_kvm_steal_time = true;
2602 }
2603 DEFINE_VIRT_MACHINE(5, 1)
2604 
2605 static void virt_machine_5_0_options(MachineClass *mc)
2606 {
2607     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2608 
2609     virt_machine_5_1_options(mc);
2610     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
2611     mc->numa_mem_supported = true;
2612     vmc->acpi_expose_flash = true;
2613     mc->auto_enable_numa_with_memdev = false;
2614 }
2615 DEFINE_VIRT_MACHINE(5, 0)
2616 
2617 static void virt_machine_4_2_options(MachineClass *mc)
2618 {
2619     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2620 
2621     virt_machine_5_0_options(mc);
2622     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
2623     vmc->kvm_no_adjvtime = true;
2624 }
2625 DEFINE_VIRT_MACHINE(4, 2)
2626 
2627 static void virt_machine_4_1_options(MachineClass *mc)
2628 {
2629     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2630 
2631     virt_machine_4_2_options(mc);
2632     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2633     vmc->no_ged = true;
2634     mc->auto_enable_numa_with_memhp = false;
2635 }
2636 DEFINE_VIRT_MACHINE(4, 1)
2637 
2638 static void virt_machine_4_0_options(MachineClass *mc)
2639 {
2640     virt_machine_4_1_options(mc);
2641     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2642 }
2643 DEFINE_VIRT_MACHINE(4, 0)
2644 
2645 static void virt_machine_3_1_options(MachineClass *mc)
2646 {
2647     virt_machine_4_0_options(mc);
2648     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2649 }
2650 DEFINE_VIRT_MACHINE(3, 1)
2651 
2652 static void virt_machine_3_0_options(MachineClass *mc)
2653 {
2654     virt_machine_3_1_options(mc);
2655     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2656 }
2657 DEFINE_VIRT_MACHINE(3, 0)
2658 
2659 static void virt_machine_2_12_options(MachineClass *mc)
2660 {
2661     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2662 
2663     virt_machine_3_0_options(mc);
2664     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2665     vmc->no_highmem_ecam = true;
2666     mc->max_cpus = 255;
2667 }
2668 DEFINE_VIRT_MACHINE(2, 12)
2669 
2670 static void virt_machine_2_11_options(MachineClass *mc)
2671 {
2672     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2673 
2674     virt_machine_2_12_options(mc);
2675     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2676     vmc->smbios_old_sys_ver = true;
2677 }
2678 DEFINE_VIRT_MACHINE(2, 11)
2679 
2680 static void virt_machine_2_10_options(MachineClass *mc)
2681 {
2682     virt_machine_2_11_options(mc);
2683     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2684     /* before 2.11 we never faulted accesses to bad addresses */
2685     mc->ignore_memory_transaction_failures = true;
2686 }
2687 DEFINE_VIRT_MACHINE(2, 10)
2688 
2689 static void virt_machine_2_9_options(MachineClass *mc)
2690 {
2691     virt_machine_2_10_options(mc);
2692     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2693 }
2694 DEFINE_VIRT_MACHINE(2, 9)
2695 
2696 static void virt_machine_2_8_options(MachineClass *mc)
2697 {
2698     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2699 
2700     virt_machine_2_9_options(mc);
2701     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
2702     /* For 2.8 and earlier we falsely claimed in the DT that
2703      * our timers were edge-triggered, not level-triggered.
2704      */
2705     vmc->claim_edge_triggered_timers = true;
2706 }
2707 DEFINE_VIRT_MACHINE(2, 8)
2708 
2709 static void virt_machine_2_7_options(MachineClass *mc)
2710 {
2711     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2712 
2713     virt_machine_2_8_options(mc);
2714     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2715     /* ITS was introduced with 2.8 */
2716     vmc->no_its = true;
2717     /* Stick with 1K pages for migration compatibility */
2718     mc->minimum_page_bits = 0;
2719 }
2720 DEFINE_VIRT_MACHINE(2, 7)
2721 
2722 static void virt_machine_2_6_options(MachineClass *mc)
2723 {
2724     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2725 
2726     virt_machine_2_7_options(mc);
2727     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
2728     vmc->disallow_affinity_adjustment = true;
2729     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2730     vmc->no_pmu = true;
2731 }
2732 DEFINE_VIRT_MACHINE(2, 6)
2733