xref: /openbmc/qemu/hw/arm/virt.c (revision fd990e86)
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/kvm.h"
52 #include "hw/loader.h"
53 #include "exec/address-spaces.h"
54 #include "qemu/bitops.h"
55 #include "qemu/error-report.h"
56 #include "qemu/module.h"
57 #include "hw/pci-host/gpex.h"
58 #include "hw/virtio/virtio-pci.h"
59 #include "hw/arm/sysbus-fdt.h"
60 #include "hw/platform-bus.h"
61 #include "hw/qdev-properties.h"
62 #include "hw/arm/fdt.h"
63 #include "hw/intc/arm_gic.h"
64 #include "hw/intc/arm_gicv3_common.h"
65 #include "hw/irq.h"
66 #include "kvm_arm.h"
67 #include "hw/firmware/smbios.h"
68 #include "qapi/visitor.h"
69 #include "standard-headers/linux/input.h"
70 #include "hw/arm/smmuv3.h"
71 #include "hw/acpi/acpi.h"
72 #include "target/arm/internals.h"
73 #include "hw/mem/pc-dimm.h"
74 #include "hw/mem/nvdimm.h"
75 #include "hw/acpi/generic_event_device.h"
76 #include "hw/virtio/virtio-iommu.h"
77 #include "hw/char/pl011.h"
78 
79 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
80     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
81                                                     void *data) \
82     { \
83         MachineClass *mc = MACHINE_CLASS(oc); \
84         virt_machine_##major##_##minor##_options(mc); \
85         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
86         if (latest) { \
87             mc->alias = "virt"; \
88         } \
89     } \
90     static const TypeInfo machvirt_##major##_##minor##_info = { \
91         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
92         .parent = TYPE_VIRT_MACHINE, \
93         .class_init = virt_##major##_##minor##_class_init, \
94     }; \
95     static void machvirt_machine_##major##_##minor##_init(void) \
96     { \
97         type_register_static(&machvirt_##major##_##minor##_info); \
98     } \
99     type_init(machvirt_machine_##major##_##minor##_init);
100 
101 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
102     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
103 #define DEFINE_VIRT_MACHINE(major, minor) \
104     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
105 
106 
107 /* Number of external interrupt lines to configure the GIC with */
108 #define NUM_IRQS 256
109 
110 #define PLATFORM_BUS_NUM_IRQS 64
111 
112 /* Legacy RAM limit in GB (< version 4.0) */
113 #define LEGACY_RAMLIMIT_GB 255
114 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
115 
116 /* Addresses and sizes of our components.
117  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
118  * 128MB..256MB is used for miscellaneous device I/O.
119  * 256MB..1GB is reserved for possible future PCI support (ie where the
120  * PCI memory window will go if we add a PCI host controller).
121  * 1GB and up is RAM (which may happily spill over into the
122  * high memory region beyond 4GB).
123  * This represents a compromise between how much RAM can be given to
124  * a 32 bit VM and leaving space for expansion and in particular for PCI.
125  * Note that devices should generally be placed at multiples of 0x10000,
126  * to accommodate guests using 64K pages.
127  */
128 static const MemMapEntry base_memmap[] = {
129     /* Space up to 0x8000000 is reserved for a boot ROM */
130     [VIRT_FLASH] =              {          0, 0x08000000 },
131     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
132     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
133     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
134     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
135     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
136     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
137     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
138     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
139     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
140     /* This redistributor space allows up to 2*64kB*123 CPUs */
141     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
142     [VIRT_UART] =               { 0x09000000, 0x00001000 },
143     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
144     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
145     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
146     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
147     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
148     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
149     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
150     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
151     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
152     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
153     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
154     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
155     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
156     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
157     /* Actual RAM size depends on initial RAM and device memory settings */
158     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
159 };
160 
161 /*
162  * Highmem IO Regions: This memory map is floating, located after the RAM.
163  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
164  * top of the RAM, so that its base get the same alignment as the size,
165  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
166  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
167  * Note the extended_memmap is sized so that it eventually also includes the
168  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
169  * index of base_memmap).
170  */
171 static MemMapEntry extended_memmap[] = {
172     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
173     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
174     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
175     /* Second PCIe window */
176     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
177 };
178 
179 static const int a15irqmap[] = {
180     [VIRT_UART] = 1,
181     [VIRT_RTC] = 2,
182     [VIRT_PCIE] = 3, /* ... to 6 */
183     [VIRT_GPIO] = 7,
184     [VIRT_SECURE_UART] = 8,
185     [VIRT_ACPI_GED] = 9,
186     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
187     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
188     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
189     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
190 };
191 
192 static const char *valid_cpus[] = {
193     ARM_CPU_TYPE_NAME("cortex-a7"),
194     ARM_CPU_TYPE_NAME("cortex-a15"),
195     ARM_CPU_TYPE_NAME("cortex-a53"),
196     ARM_CPU_TYPE_NAME("cortex-a57"),
197     ARM_CPU_TYPE_NAME("cortex-a72"),
198     ARM_CPU_TYPE_NAME("host"),
199     ARM_CPU_TYPE_NAME("max"),
200 };
201 
202 static bool cpu_type_valid(const char *cpu)
203 {
204     int i;
205 
206     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
207         if (strcmp(cpu, valid_cpus[i]) == 0) {
208             return true;
209         }
210     }
211     return false;
212 }
213 
214 static void create_fdt(VirtMachineState *vms)
215 {
216     MachineState *ms = MACHINE(vms);
217     int nb_numa_nodes = ms->numa_state->num_nodes;
218     void *fdt = create_device_tree(&vms->fdt_size);
219 
220     if (!fdt) {
221         error_report("create_device_tree() failed");
222         exit(1);
223     }
224 
225     vms->fdt = fdt;
226 
227     /* Header */
228     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
229     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
230     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
231 
232     /* /chosen must exist for load_dtb to fill in necessary properties later */
233     qemu_fdt_add_subnode(fdt, "/chosen");
234 
235     /* Clock node, for the benefit of the UART. The kernel device tree
236      * binding documentation claims the PL011 node clock properties are
237      * optional but in practice if you omit them the kernel refuses to
238      * probe for the device.
239      */
240     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
241     qemu_fdt_add_subnode(fdt, "/apb-pclk");
242     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
243     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
244     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
245     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
246                                 "clk24mhz");
247     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
248 
249     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
250         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
251         uint32_t *matrix = g_malloc0(size);
252         int idx, i, j;
253 
254         for (i = 0; i < nb_numa_nodes; i++) {
255             for (j = 0; j < nb_numa_nodes; j++) {
256                 idx = (i * nb_numa_nodes + j) * 3;
257                 matrix[idx + 0] = cpu_to_be32(i);
258                 matrix[idx + 1] = cpu_to_be32(j);
259                 matrix[idx + 2] =
260                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
261             }
262         }
263 
264         qemu_fdt_add_subnode(fdt, "/distance-map");
265         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
266                                 "numa-distance-map-v1");
267         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
268                          matrix, size);
269         g_free(matrix);
270     }
271 }
272 
273 static void fdt_add_timer_nodes(const VirtMachineState *vms)
274 {
275     /* On real hardware these interrupts are level-triggered.
276      * On KVM they were edge-triggered before host kernel version 4.4,
277      * and level-triggered afterwards.
278      * On emulated QEMU they are level-triggered.
279      *
280      * Getting the DTB info about them wrong is awkward for some
281      * guest kernels:
282      *  pre-4.8 ignore the DT and leave the interrupt configured
283      *   with whatever the GIC reset value (or the bootloader) left it at
284      *  4.8 before rc6 honour the incorrect data by programming it back
285      *   into the GIC, causing problems
286      *  4.8rc6 and later ignore the DT and always write "level triggered"
287      *   into the GIC
288      *
289      * For backwards-compatibility, virt-2.8 and earlier will continue
290      * to say these are edge-triggered, but later machines will report
291      * the correct information.
292      */
293     ARMCPU *armcpu;
294     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
295     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
296 
297     if (vmc->claim_edge_triggered_timers) {
298         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
299     }
300 
301     if (vms->gic_version == 2) {
302         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
303                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
304                              (1 << vms->smp_cpus) - 1);
305     }
306 
307     qemu_fdt_add_subnode(vms->fdt, "/timer");
308 
309     armcpu = ARM_CPU(qemu_get_cpu(0));
310     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
311         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
312         qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
313                          compat, sizeof(compat));
314     } else {
315         qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
316                                 "arm,armv7-timer");
317     }
318     qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
319     qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
320                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
321                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
322                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
323                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
324 }
325 
326 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
327 {
328     int cpu;
329     int addr_cells = 1;
330     const MachineState *ms = MACHINE(vms);
331 
332     /*
333      * From Documentation/devicetree/bindings/arm/cpus.txt
334      *  On ARM v8 64-bit systems value should be set to 2,
335      *  that corresponds to the MPIDR_EL1 register size.
336      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
337      *  in the system, #address-cells can be set to 1, since
338      *  MPIDR_EL1[63:32] bits are not used for CPUs
339      *  identification.
340      *
341      *  Here we actually don't know whether our system is 32- or 64-bit one.
342      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
343      *  at least one of them has Aff3 populated, we set #address-cells to 2.
344      */
345     for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
346         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
347 
348         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
349             addr_cells = 2;
350             break;
351         }
352     }
353 
354     qemu_fdt_add_subnode(vms->fdt, "/cpus");
355     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
356     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
357 
358     for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
359         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
360         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
361         CPUState *cs = CPU(armcpu);
362 
363         qemu_fdt_add_subnode(vms->fdt, nodename);
364         qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
365         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
366                                     armcpu->dtb_compatible);
367 
368         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
369             && vms->smp_cpus > 1) {
370             qemu_fdt_setprop_string(vms->fdt, nodename,
371                                         "enable-method", "psci");
372         }
373 
374         if (addr_cells == 2) {
375             qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
376                                  armcpu->mp_affinity);
377         } else {
378             qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
379                                   armcpu->mp_affinity);
380         }
381 
382         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
383             qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
384                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
385         }
386 
387         g_free(nodename);
388     }
389 }
390 
391 static void fdt_add_its_gic_node(VirtMachineState *vms)
392 {
393     char *nodename;
394 
395     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
396     nodename = g_strdup_printf("/intc/its@%" PRIx64,
397                                vms->memmap[VIRT_GIC_ITS].base);
398     qemu_fdt_add_subnode(vms->fdt, nodename);
399     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
400                             "arm,gic-v3-its");
401     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
402     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
403                                  2, vms->memmap[VIRT_GIC_ITS].base,
404                                  2, vms->memmap[VIRT_GIC_ITS].size);
405     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
406     g_free(nodename);
407 }
408 
409 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
410 {
411     char *nodename;
412 
413     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
414                                vms->memmap[VIRT_GIC_V2M].base);
415     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
416     qemu_fdt_add_subnode(vms->fdt, nodename);
417     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
418                             "arm,gic-v2m-frame");
419     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
420     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
421                                  2, vms->memmap[VIRT_GIC_V2M].base,
422                                  2, vms->memmap[VIRT_GIC_V2M].size);
423     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
424     g_free(nodename);
425 }
426 
427 static void fdt_add_gic_node(VirtMachineState *vms)
428 {
429     char *nodename;
430 
431     vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
432     qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
433 
434     nodename = g_strdup_printf("/intc@%" PRIx64,
435                                vms->memmap[VIRT_GIC_DIST].base);
436     qemu_fdt_add_subnode(vms->fdt, nodename);
437     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 3);
438     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-controller", NULL, 0);
439     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 0x2);
440     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 0x2);
441     qemu_fdt_setprop(vms->fdt, nodename, "ranges", NULL, 0);
442     if (vms->gic_version == 3) {
443         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
444 
445         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
446                                 "arm,gic-v3");
447 
448         qemu_fdt_setprop_cell(vms->fdt, nodename,
449                               "#redistributor-regions", nb_redist_regions);
450 
451         if (nb_redist_regions == 1) {
452             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
453                                          2, vms->memmap[VIRT_GIC_DIST].base,
454                                          2, vms->memmap[VIRT_GIC_DIST].size,
455                                          2, vms->memmap[VIRT_GIC_REDIST].base,
456                                          2, vms->memmap[VIRT_GIC_REDIST].size);
457         } else {
458             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
459                                  2, vms->memmap[VIRT_GIC_DIST].base,
460                                  2, vms->memmap[VIRT_GIC_DIST].size,
461                                  2, vms->memmap[VIRT_GIC_REDIST].base,
462                                  2, vms->memmap[VIRT_GIC_REDIST].size,
463                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
464                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
465         }
466 
467         if (vms->virt) {
468             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
469                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
470                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
471         }
472     } else {
473         /* 'cortex-a15-gic' means 'GIC v2' */
474         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
475                                 "arm,cortex-a15-gic");
476         if (!vms->virt) {
477             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
478                                          2, vms->memmap[VIRT_GIC_DIST].base,
479                                          2, vms->memmap[VIRT_GIC_DIST].size,
480                                          2, vms->memmap[VIRT_GIC_CPU].base,
481                                          2, vms->memmap[VIRT_GIC_CPU].size);
482         } else {
483             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
484                                          2, vms->memmap[VIRT_GIC_DIST].base,
485                                          2, vms->memmap[VIRT_GIC_DIST].size,
486                                          2, vms->memmap[VIRT_GIC_CPU].base,
487                                          2, vms->memmap[VIRT_GIC_CPU].size,
488                                          2, vms->memmap[VIRT_GIC_HYP].base,
489                                          2, vms->memmap[VIRT_GIC_HYP].size,
490                                          2, vms->memmap[VIRT_GIC_VCPU].base,
491                                          2, vms->memmap[VIRT_GIC_VCPU].size);
492             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
493                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
494                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
495         }
496     }
497 
498     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->gic_phandle);
499     g_free(nodename);
500 }
501 
502 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
503 {
504     CPUState *cpu;
505     ARMCPU *armcpu;
506     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
507 
508     CPU_FOREACH(cpu) {
509         armcpu = ARM_CPU(cpu);
510         if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
511             return;
512         }
513         if (kvm_enabled()) {
514             if (kvm_irqchip_in_kernel()) {
515                 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
516             }
517             kvm_arm_pmu_init(cpu);
518         }
519     }
520 
521     if (vms->gic_version == 2) {
522         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
523                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
524                              (1 << vms->smp_cpus) - 1);
525     }
526 
527     armcpu = ARM_CPU(qemu_get_cpu(0));
528     qemu_fdt_add_subnode(vms->fdt, "/pmu");
529     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
530         const char compat[] = "arm,armv8-pmuv3";
531         qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
532                          compat, sizeof(compat));
533         qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
534                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
535     }
536 }
537 
538 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
539 {
540     DeviceState *dev;
541     MachineState *ms = MACHINE(vms);
542     int irq = vms->irqmap[VIRT_ACPI_GED];
543     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
544 
545     if (ms->ram_slots) {
546         event |= ACPI_GED_MEM_HOTPLUG_EVT;
547     }
548 
549     dev = qdev_create(NULL, TYPE_ACPI_GED);
550     qdev_prop_set_uint32(dev, "ged-event", event);
551 
552     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
553     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
554     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
555 
556     qdev_init_nofail(dev);
557 
558     return dev;
559 }
560 
561 static void create_its(VirtMachineState *vms)
562 {
563     const char *itsclass = its_class_name();
564     DeviceState *dev;
565 
566     if (!itsclass) {
567         /* Do nothing if not supported */
568         return;
569     }
570 
571     dev = qdev_create(NULL, itsclass);
572 
573     object_property_set_link(OBJECT(dev), OBJECT(vms->gic), "parent-gicv3",
574                              &error_abort);
575     qdev_init_nofail(dev);
576     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
577 
578     fdt_add_its_gic_node(vms);
579 }
580 
581 static void create_v2m(VirtMachineState *vms)
582 {
583     int i;
584     int irq = vms->irqmap[VIRT_GIC_V2M];
585     DeviceState *dev;
586 
587     dev = qdev_create(NULL, "arm-gicv2m");
588     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
589     qdev_prop_set_uint32(dev, "base-spi", irq);
590     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
591     qdev_init_nofail(dev);
592 
593     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
594         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
595                            qdev_get_gpio_in(vms->gic, irq + i));
596     }
597 
598     fdt_add_v2m_gic_node(vms);
599 }
600 
601 static void create_gic(VirtMachineState *vms)
602 {
603     MachineState *ms = MACHINE(vms);
604     /* We create a standalone GIC */
605     SysBusDevice *gicbusdev;
606     const char *gictype;
607     int type = vms->gic_version, i;
608     unsigned int smp_cpus = ms->smp.cpus;
609     uint32_t nb_redist_regions = 0;
610 
611     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
612 
613     vms->gic = qdev_create(NULL, gictype);
614     qdev_prop_set_uint32(vms->gic, "revision", type);
615     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
616     /* Note that the num-irq property counts both internal and external
617      * interrupts; there are always 32 of the former (mandated by GIC spec).
618      */
619     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
620     if (!kvm_irqchip_in_kernel()) {
621         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
622     }
623 
624     if (type == 3) {
625         uint32_t redist0_capacity =
626                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
627         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
628 
629         nb_redist_regions = virt_gicv3_redist_region_count(vms);
630 
631         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
632                              nb_redist_regions);
633         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
634 
635         if (nb_redist_regions == 2) {
636             uint32_t redist1_capacity =
637                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
638 
639             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
640                 MIN(smp_cpus - redist0_count, redist1_capacity));
641         }
642     } else {
643         if (!kvm_irqchip_in_kernel()) {
644             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
645                               vms->virt);
646         }
647     }
648     qdev_init_nofail(vms->gic);
649     gicbusdev = SYS_BUS_DEVICE(vms->gic);
650     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
651     if (type == 3) {
652         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
653         if (nb_redist_regions == 2) {
654             sysbus_mmio_map(gicbusdev, 2,
655                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
656         }
657     } else {
658         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
659         if (vms->virt) {
660             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
661             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
662         }
663     }
664 
665     /* Wire the outputs from each CPU's generic timer and the GICv3
666      * maintenance interrupt signal to the appropriate GIC PPI inputs,
667      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
668      */
669     for (i = 0; i < smp_cpus; i++) {
670         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
671         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
672         int irq;
673         /* Mapping from the output timer irq lines from the CPU to the
674          * GIC PPI inputs we use for the virt board.
675          */
676         const int timer_irq[] = {
677             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
678             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
679             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
680             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
681         };
682 
683         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
684             qdev_connect_gpio_out(cpudev, irq,
685                                   qdev_get_gpio_in(vms->gic,
686                                                    ppibase + timer_irq[irq]));
687         }
688 
689         if (type == 3) {
690             qemu_irq irq = qdev_get_gpio_in(vms->gic,
691                                             ppibase + ARCH_GIC_MAINT_IRQ);
692             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
693                                         0, irq);
694         } else if (vms->virt) {
695             qemu_irq irq = qdev_get_gpio_in(vms->gic,
696                                             ppibase + ARCH_GIC_MAINT_IRQ);
697             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
698         }
699 
700         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
701                                     qdev_get_gpio_in(vms->gic, ppibase
702                                                      + VIRTUAL_PMU_IRQ));
703 
704         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
705         sysbus_connect_irq(gicbusdev, i + smp_cpus,
706                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
707         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
708                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
709         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
710                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
711     }
712 
713     fdt_add_gic_node(vms);
714 
715     if (type == 3 && vms->its) {
716         create_its(vms);
717     } else if (type == 2) {
718         create_v2m(vms);
719     }
720 }
721 
722 static void create_uart(const VirtMachineState *vms, int uart,
723                         MemoryRegion *mem, Chardev *chr)
724 {
725     char *nodename;
726     hwaddr base = vms->memmap[uart].base;
727     hwaddr size = vms->memmap[uart].size;
728     int irq = vms->irqmap[uart];
729     const char compat[] = "arm,pl011\0arm,primecell";
730     const char clocknames[] = "uartclk\0apb_pclk";
731     DeviceState *dev = qdev_create(NULL, TYPE_PL011);
732     SysBusDevice *s = SYS_BUS_DEVICE(dev);
733 
734     qdev_prop_set_chr(dev, "chardev", chr);
735     qdev_init_nofail(dev);
736     memory_region_add_subregion(mem, base,
737                                 sysbus_mmio_get_region(s, 0));
738     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
739 
740     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
741     qemu_fdt_add_subnode(vms->fdt, nodename);
742     /* Note that we can't use setprop_string because of the embedded NUL */
743     qemu_fdt_setprop(vms->fdt, nodename, "compatible",
744                          compat, sizeof(compat));
745     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
746                                      2, base, 2, size);
747     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
748                                GIC_FDT_IRQ_TYPE_SPI, irq,
749                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
750     qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
751                                vms->clock_phandle, vms->clock_phandle);
752     qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
753                          clocknames, sizeof(clocknames));
754 
755     if (uart == VIRT_UART) {
756         qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
757     } else {
758         /* Mark as not usable by the normal world */
759         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
760         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
761 
762         qemu_fdt_add_subnode(vms->fdt, "/secure-chosen");
763         qemu_fdt_setprop_string(vms->fdt, "/secure-chosen", "stdout-path",
764                                 nodename);
765     }
766 
767     g_free(nodename);
768 }
769 
770 static void create_rtc(const VirtMachineState *vms)
771 {
772     char *nodename;
773     hwaddr base = vms->memmap[VIRT_RTC].base;
774     hwaddr size = vms->memmap[VIRT_RTC].size;
775     int irq = vms->irqmap[VIRT_RTC];
776     const char compat[] = "arm,pl031\0arm,primecell";
777 
778     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
779 
780     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
781     qemu_fdt_add_subnode(vms->fdt, nodename);
782     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
783     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
784                                  2, base, 2, size);
785     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
786                            GIC_FDT_IRQ_TYPE_SPI, irq,
787                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
788     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
789     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
790     g_free(nodename);
791 }
792 
793 static DeviceState *gpio_key_dev;
794 static void virt_powerdown_req(Notifier *n, void *opaque)
795 {
796     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
797 
798     if (s->acpi_dev) {
799         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
800     } else {
801         /* use gpio Pin 3 for power button event */
802         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
803     }
804 }
805 
806 static void create_gpio(const VirtMachineState *vms)
807 {
808     char *nodename;
809     DeviceState *pl061_dev;
810     hwaddr base = vms->memmap[VIRT_GPIO].base;
811     hwaddr size = vms->memmap[VIRT_GPIO].size;
812     int irq = vms->irqmap[VIRT_GPIO];
813     const char compat[] = "arm,pl061\0arm,primecell";
814 
815     pl061_dev = sysbus_create_simple("pl061", base,
816                                      qdev_get_gpio_in(vms->gic, irq));
817 
818     uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
819     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
820     qemu_fdt_add_subnode(vms->fdt, nodename);
821     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
822                                  2, base, 2, size);
823     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
824     qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
825     qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
826     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
827                            GIC_FDT_IRQ_TYPE_SPI, irq,
828                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
829     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
830     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
831     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
832 
833     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
834                                         qdev_get_gpio_in(pl061_dev, 3));
835     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
836     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
837     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
838     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
839 
840     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
841     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
842                             "label", "GPIO Key Poweroff");
843     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
844                           KEY_POWER);
845     qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
846                            "gpios", phandle, 3, 0);
847     g_free(nodename);
848 }
849 
850 static void create_virtio_devices(const VirtMachineState *vms)
851 {
852     int i;
853     hwaddr size = vms->memmap[VIRT_MMIO].size;
854 
855     /* We create the transports in forwards order. Since qbus_realize()
856      * prepends (not appends) new child buses, the incrementing loop below will
857      * create a list of virtio-mmio buses with decreasing base addresses.
858      *
859      * When a -device option is processed from the command line,
860      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
861      * order. The upshot is that -device options in increasing command line
862      * order are mapped to virtio-mmio buses with decreasing base addresses.
863      *
864      * When this code was originally written, that arrangement ensured that the
865      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
866      * the first -device on the command line. (The end-to-end order is a
867      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
868      * guest kernel's name-to-address assignment strategy.)
869      *
870      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
871      * the message, if not necessarily the code, of commit 70161ff336.
872      * Therefore the loop now establishes the inverse of the original intent.
873      *
874      * Unfortunately, we can't counteract the kernel change by reversing the
875      * loop; it would break existing command lines.
876      *
877      * In any case, the kernel makes no guarantee about the stability of
878      * enumeration order of virtio devices (as demonstrated by it changing
879      * between kernel versions). For reliable and stable identification
880      * of disks users must use UUIDs or similar mechanisms.
881      */
882     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
883         int irq = vms->irqmap[VIRT_MMIO] + i;
884         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
885 
886         sysbus_create_simple("virtio-mmio", base,
887                              qdev_get_gpio_in(vms->gic, irq));
888     }
889 
890     /* We add dtb nodes in reverse order so that they appear in the finished
891      * device tree lowest address first.
892      *
893      * Note that this mapping is independent of the loop above. The previous
894      * loop influences virtio device to virtio transport assignment, whereas
895      * this loop controls how virtio transports are laid out in the dtb.
896      */
897     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
898         char *nodename;
899         int irq = vms->irqmap[VIRT_MMIO] + i;
900         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
901 
902         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
903         qemu_fdt_add_subnode(vms->fdt, nodename);
904         qemu_fdt_setprop_string(vms->fdt, nodename,
905                                 "compatible", "virtio,mmio");
906         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
907                                      2, base, 2, size);
908         qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
909                                GIC_FDT_IRQ_TYPE_SPI, irq,
910                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
911         qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
912         g_free(nodename);
913     }
914 }
915 
916 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
917 
918 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
919                                         const char *name,
920                                         const char *alias_prop_name)
921 {
922     /*
923      * Create a single flash device.  We use the same parameters as
924      * the flash devices on the Versatile Express board.
925      */
926     DeviceState *dev = qdev_create(NULL, TYPE_PFLASH_CFI01);
927 
928     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
929     qdev_prop_set_uint8(dev, "width", 4);
930     qdev_prop_set_uint8(dev, "device-width", 2);
931     qdev_prop_set_bit(dev, "big-endian", false);
932     qdev_prop_set_uint16(dev, "id0", 0x89);
933     qdev_prop_set_uint16(dev, "id1", 0x18);
934     qdev_prop_set_uint16(dev, "id2", 0x00);
935     qdev_prop_set_uint16(dev, "id3", 0x00);
936     qdev_prop_set_string(dev, "name", name);
937     object_property_add_child(OBJECT(vms), name, OBJECT(dev),
938                               &error_abort);
939     object_property_add_alias(OBJECT(vms), alias_prop_name,
940                               OBJECT(dev), "drive", &error_abort);
941     return PFLASH_CFI01(dev);
942 }
943 
944 static void virt_flash_create(VirtMachineState *vms)
945 {
946     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
947     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
948 }
949 
950 static void virt_flash_map1(PFlashCFI01 *flash,
951                             hwaddr base, hwaddr size,
952                             MemoryRegion *sysmem)
953 {
954     DeviceState *dev = DEVICE(flash);
955 
956     assert(size % VIRT_FLASH_SECTOR_SIZE == 0);
957     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
958     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
959     qdev_init_nofail(dev);
960 
961     memory_region_add_subregion(sysmem, base,
962                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
963                                                        0));
964 }
965 
966 static void virt_flash_map(VirtMachineState *vms,
967                            MemoryRegion *sysmem,
968                            MemoryRegion *secure_sysmem)
969 {
970     /*
971      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
972      * sysmem is the system memory space. secure_sysmem is the secure view
973      * of the system, and the first flash device should be made visible only
974      * there. The second flash device is visible to both secure and nonsecure.
975      * If sysmem == secure_sysmem this means there is no separate Secure
976      * address space and both flash devices are generally visible.
977      */
978     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
979     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
980 
981     virt_flash_map1(vms->flash[0], flashbase, flashsize,
982                     secure_sysmem);
983     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
984                     sysmem);
985 }
986 
987 static void virt_flash_fdt(VirtMachineState *vms,
988                            MemoryRegion *sysmem,
989                            MemoryRegion *secure_sysmem)
990 {
991     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
992     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
993     char *nodename;
994 
995     if (sysmem == secure_sysmem) {
996         /* Report both flash devices as a single node in the DT */
997         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
998         qemu_fdt_add_subnode(vms->fdt, nodename);
999         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1000         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1001                                      2, flashbase, 2, flashsize,
1002                                      2, flashbase + flashsize, 2, flashsize);
1003         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1004         g_free(nodename);
1005     } else {
1006         /*
1007          * Report the devices as separate nodes so we can mark one as
1008          * only visible to the secure world.
1009          */
1010         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1011         qemu_fdt_add_subnode(vms->fdt, nodename);
1012         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1013         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1014                                      2, flashbase, 2, flashsize);
1015         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1016         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1017         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1018         g_free(nodename);
1019 
1020         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1021         qemu_fdt_add_subnode(vms->fdt, nodename);
1022         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1023         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1024                                      2, flashbase + flashsize, 2, flashsize);
1025         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1026         g_free(nodename);
1027     }
1028 }
1029 
1030 static bool virt_firmware_init(VirtMachineState *vms,
1031                                MemoryRegion *sysmem,
1032                                MemoryRegion *secure_sysmem)
1033 {
1034     int i;
1035     BlockBackend *pflash_blk0;
1036 
1037     /* Map legacy -drive if=pflash to machine properties */
1038     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1039         pflash_cfi01_legacy_drive(vms->flash[i],
1040                                   drive_get(IF_PFLASH, 0, i));
1041     }
1042 
1043     virt_flash_map(vms, sysmem, secure_sysmem);
1044 
1045     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1046 
1047     if (bios_name) {
1048         char *fname;
1049         MemoryRegion *mr;
1050         int image_size;
1051 
1052         if (pflash_blk0) {
1053             error_report("The contents of the first flash device may be "
1054                          "specified with -bios or with -drive if=pflash... "
1055                          "but you cannot use both options at once");
1056             exit(1);
1057         }
1058 
1059         /* Fall back to -bios */
1060 
1061         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1062         if (!fname) {
1063             error_report("Could not find ROM image '%s'", bios_name);
1064             exit(1);
1065         }
1066         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1067         image_size = load_image_mr(fname, mr);
1068         g_free(fname);
1069         if (image_size < 0) {
1070             error_report("Could not load ROM image '%s'", bios_name);
1071             exit(1);
1072         }
1073     }
1074 
1075     return pflash_blk0 || bios_name;
1076 }
1077 
1078 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1079 {
1080     MachineState *ms = MACHINE(vms);
1081     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1082     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1083     FWCfgState *fw_cfg;
1084     char *nodename;
1085 
1086     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1087     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1088 
1089     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1090     qemu_fdt_add_subnode(vms->fdt, nodename);
1091     qemu_fdt_setprop_string(vms->fdt, nodename,
1092                             "compatible", "qemu,fw-cfg-mmio");
1093     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1094                                  2, base, 2, size);
1095     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1096     g_free(nodename);
1097     return fw_cfg;
1098 }
1099 
1100 static void create_pcie_irq_map(const VirtMachineState *vms,
1101                                 uint32_t gic_phandle,
1102                                 int first_irq, const char *nodename)
1103 {
1104     int devfn, pin;
1105     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1106     uint32_t *irq_map = full_irq_map;
1107 
1108     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1109         for (pin = 0; pin < 4; pin++) {
1110             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1111             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1112             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1113             int i;
1114 
1115             uint32_t map[] = {
1116                 devfn << 8, 0, 0,                           /* devfn */
1117                 pin + 1,                                    /* PCI pin */
1118                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1119 
1120             /* Convert map to big endian */
1121             for (i = 0; i < 10; i++) {
1122                 irq_map[i] = cpu_to_be32(map[i]);
1123             }
1124             irq_map += 10;
1125         }
1126     }
1127 
1128     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
1129                      full_irq_map, sizeof(full_irq_map));
1130 
1131     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
1132                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
1133                            0x7           /* PCI irq */);
1134 }
1135 
1136 static void create_smmu(const VirtMachineState *vms,
1137                         PCIBus *bus)
1138 {
1139     char *node;
1140     const char compat[] = "arm,smmu-v3";
1141     int irq =  vms->irqmap[VIRT_SMMU];
1142     int i;
1143     hwaddr base = vms->memmap[VIRT_SMMU].base;
1144     hwaddr size = vms->memmap[VIRT_SMMU].size;
1145     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1146     DeviceState *dev;
1147 
1148     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1149         return;
1150     }
1151 
1152     dev = qdev_create(NULL, "arm-smmuv3");
1153 
1154     object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
1155                              &error_abort);
1156     qdev_init_nofail(dev);
1157     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1158     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1159         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1160                            qdev_get_gpio_in(vms->gic, irq + i));
1161     }
1162 
1163     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1164     qemu_fdt_add_subnode(vms->fdt, node);
1165     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1166     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
1167 
1168     qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
1169             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1170             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1171             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1172             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1173 
1174     qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
1175                      sizeof(irq_names));
1176 
1177     qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
1178     qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
1179     qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
1180 
1181     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1182 
1183     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1184     g_free(node);
1185 }
1186 
1187 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms, Error **errp)
1188 {
1189     const char compat[] = "virtio,pci-iommu";
1190     uint16_t bdf = vms->virtio_iommu_bdf;
1191     char *node;
1192 
1193     vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1194 
1195     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1196     qemu_fdt_add_subnode(vms->fdt, node);
1197     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1198     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg",
1199                                  1, bdf << 8, 1, 0, 1, 0,
1200                                  1, 0, 1, 0);
1201 
1202     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1203     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1204     g_free(node);
1205 
1206     qemu_fdt_setprop_cells(vms->fdt, vms->pciehb_nodename, "iommu-map",
1207                            0x0, vms->iommu_phandle, 0x0, bdf,
1208                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1209 }
1210 
1211 static void create_pcie(VirtMachineState *vms)
1212 {
1213     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1214     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1215     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1216     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1217     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1218     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1219     hwaddr base_ecam, size_ecam;
1220     hwaddr base = base_mmio;
1221     int nr_pcie_buses;
1222     int irq = vms->irqmap[VIRT_PCIE];
1223     MemoryRegion *mmio_alias;
1224     MemoryRegion *mmio_reg;
1225     MemoryRegion *ecam_alias;
1226     MemoryRegion *ecam_reg;
1227     DeviceState *dev;
1228     char *nodename;
1229     int i, ecam_id;
1230     PCIHostState *pci;
1231 
1232     dev = qdev_create(NULL, TYPE_GPEX_HOST);
1233     qdev_init_nofail(dev);
1234 
1235     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1236     base_ecam = vms->memmap[ecam_id].base;
1237     size_ecam = vms->memmap[ecam_id].size;
1238     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1239     /* Map only the first size_ecam bytes of ECAM space */
1240     ecam_alias = g_new0(MemoryRegion, 1);
1241     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1242     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1243                              ecam_reg, 0, size_ecam);
1244     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1245 
1246     /* Map the MMIO window into system address space so as to expose
1247      * the section of PCI MMIO space which starts at the same base address
1248      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1249      * the window).
1250      */
1251     mmio_alias = g_new0(MemoryRegion, 1);
1252     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1253     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1254                              mmio_reg, base_mmio, size_mmio);
1255     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1256 
1257     if (vms->highmem) {
1258         /* Map high MMIO space */
1259         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1260 
1261         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1262                                  mmio_reg, base_mmio_high, size_mmio_high);
1263         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1264                                     high_mmio_alias);
1265     }
1266 
1267     /* Map IO port space */
1268     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1269 
1270     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1271         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1272                            qdev_get_gpio_in(vms->gic, irq + i));
1273         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1274     }
1275 
1276     pci = PCI_HOST_BRIDGE(dev);
1277     if (pci->bus) {
1278         for (i = 0; i < nb_nics; i++) {
1279             NICInfo *nd = &nd_table[i];
1280 
1281             if (!nd->model) {
1282                 nd->model = g_strdup("virtio");
1283             }
1284 
1285             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1286         }
1287     }
1288 
1289     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1290     qemu_fdt_add_subnode(vms->fdt, nodename);
1291     qemu_fdt_setprop_string(vms->fdt, nodename,
1292                             "compatible", "pci-host-ecam-generic");
1293     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1294     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1295     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1296     qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1297     qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1298                            nr_pcie_buses - 1);
1299     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1300 
1301     if (vms->msi_phandle) {
1302         qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1303                                vms->msi_phandle);
1304     }
1305 
1306     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1307                                  2, base_ecam, 2, size_ecam);
1308 
1309     if (vms->highmem) {
1310         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1311                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1312                                      2, base_pio, 2, size_pio,
1313                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1314                                      2, base_mmio, 2, size_mmio,
1315                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1316                                      2, base_mmio_high,
1317                                      2, base_mmio_high, 2, size_mmio_high);
1318     } else {
1319         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1320                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1321                                      2, base_pio, 2, size_pio,
1322                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1323                                      2, base_mmio, 2, size_mmio);
1324     }
1325 
1326     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1327     create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1328 
1329     if (vms->iommu) {
1330         vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1331 
1332         switch (vms->iommu) {
1333         case VIRT_IOMMU_SMMUV3:
1334             create_smmu(vms, pci->bus);
1335             qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1336                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1337             break;
1338         default:
1339             g_assert_not_reached();
1340         }
1341     }
1342 }
1343 
1344 static void create_platform_bus(VirtMachineState *vms)
1345 {
1346     DeviceState *dev;
1347     SysBusDevice *s;
1348     int i;
1349     MemoryRegion *sysmem = get_system_memory();
1350 
1351     dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1352     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1353     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1354     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1355     qdev_init_nofail(dev);
1356     vms->platform_bus_dev = dev;
1357 
1358     s = SYS_BUS_DEVICE(dev);
1359     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1360         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1361         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1362     }
1363 
1364     memory_region_add_subregion(sysmem,
1365                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1366                                 sysbus_mmio_get_region(s, 0));
1367 }
1368 
1369 static void create_secure_ram(VirtMachineState *vms,
1370                               MemoryRegion *secure_sysmem)
1371 {
1372     MemoryRegion *secram = g_new(MemoryRegion, 1);
1373     char *nodename;
1374     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1375     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1376 
1377     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1378                            &error_fatal);
1379     memory_region_add_subregion(secure_sysmem, base, secram);
1380 
1381     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1382     qemu_fdt_add_subnode(vms->fdt, nodename);
1383     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1384     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1385     qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1386     qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1387 
1388     g_free(nodename);
1389 }
1390 
1391 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1392 {
1393     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1394                                                  bootinfo);
1395 
1396     *fdt_size = board->fdt_size;
1397     return board->fdt;
1398 }
1399 
1400 static void virt_build_smbios(VirtMachineState *vms)
1401 {
1402     MachineClass *mc = MACHINE_GET_CLASS(vms);
1403     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1404     uint8_t *smbios_tables, *smbios_anchor;
1405     size_t smbios_tables_len, smbios_anchor_len;
1406     const char *product = "QEMU Virtual Machine";
1407 
1408     if (kvm_enabled()) {
1409         product = "KVM Virtual Machine";
1410     }
1411 
1412     smbios_set_defaults("QEMU", product,
1413                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1414                         true, SMBIOS_ENTRY_POINT_30);
1415 
1416     smbios_get_tables(MACHINE(vms), NULL, 0, &smbios_tables, &smbios_tables_len,
1417                       &smbios_anchor, &smbios_anchor_len);
1418 
1419     if (smbios_anchor) {
1420         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1421                         smbios_tables, smbios_tables_len);
1422         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1423                         smbios_anchor, smbios_anchor_len);
1424     }
1425 }
1426 
1427 static
1428 void virt_machine_done(Notifier *notifier, void *data)
1429 {
1430     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1431                                          machine_done);
1432     MachineState *ms = MACHINE(vms);
1433     ARMCPU *cpu = ARM_CPU(first_cpu);
1434     struct arm_boot_info *info = &vms->bootinfo;
1435     AddressSpace *as = arm_boot_address_space(cpu, info);
1436 
1437     /*
1438      * If the user provided a dtb, we assume the dynamic sysbus nodes
1439      * already are integrated there. This corresponds to a use case where
1440      * the dynamic sysbus nodes are complex and their generation is not yet
1441      * supported. In that case the user can take charge of the guest dt
1442      * while qemu takes charge of the qom stuff.
1443      */
1444     if (info->dtb_filename == NULL) {
1445         platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1446                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1447                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1448                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1449     }
1450     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1451         exit(1);
1452     }
1453 
1454     virt_acpi_setup(vms);
1455     virt_build_smbios(vms);
1456 }
1457 
1458 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1459 {
1460     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1461     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1462 
1463     if (!vmc->disallow_affinity_adjustment) {
1464         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1465          * GIC's target-list limitations. 32-bit KVM hosts currently
1466          * always create clusters of 4 CPUs, but that is expected to
1467          * change when they gain support for gicv3. When KVM is enabled
1468          * it will override the changes we make here, therefore our
1469          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1470          * and to improve SGI efficiency.
1471          */
1472         if (vms->gic_version == 3) {
1473             clustersz = GICV3_TARGETLIST_BITS;
1474         } else {
1475             clustersz = GIC_TARGETLIST_BITS;
1476         }
1477     }
1478     return arm_cpu_mp_affinity(idx, clustersz);
1479 }
1480 
1481 static void virt_set_memmap(VirtMachineState *vms)
1482 {
1483     MachineState *ms = MACHINE(vms);
1484     hwaddr base, device_memory_base, device_memory_size;
1485     int i;
1486 
1487     vms->memmap = extended_memmap;
1488 
1489     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1490         vms->memmap[i] = base_memmap[i];
1491     }
1492 
1493     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1494         error_report("unsupported number of memory slots: %"PRIu64,
1495                      ms->ram_slots);
1496         exit(EXIT_FAILURE);
1497     }
1498 
1499     /*
1500      * We compute the base of the high IO region depending on the
1501      * amount of initial and device memory. The device memory start/size
1502      * is aligned on 1GiB. We never put the high IO region below 256GiB
1503      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1504      * The device region size assumes 1GiB page max alignment per slot.
1505      */
1506     device_memory_base =
1507         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1508     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1509 
1510     /* Base address of the high IO region */
1511     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1512     if (base < device_memory_base) {
1513         error_report("maxmem/slots too huge");
1514         exit(EXIT_FAILURE);
1515     }
1516     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1517         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1518     }
1519 
1520     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1521         hwaddr size = extended_memmap[i].size;
1522 
1523         base = ROUND_UP(base, size);
1524         vms->memmap[i].base = base;
1525         vms->memmap[i].size = size;
1526         base += size;
1527     }
1528     vms->highest_gpa = base - 1;
1529     if (device_memory_size > 0) {
1530         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1531         ms->device_memory->base = device_memory_base;
1532         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1533                            "device-memory", device_memory_size);
1534     }
1535 }
1536 
1537 static void machvirt_init(MachineState *machine)
1538 {
1539     VirtMachineState *vms = VIRT_MACHINE(machine);
1540     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1541     MachineClass *mc = MACHINE_GET_CLASS(machine);
1542     const CPUArchIdList *possible_cpus;
1543     MemoryRegion *sysmem = get_system_memory();
1544     MemoryRegion *secure_sysmem = NULL;
1545     int n, virt_max_cpus;
1546     bool firmware_loaded;
1547     bool aarch64 = true;
1548     bool has_ged = !vmc->no_ged;
1549     unsigned int smp_cpus = machine->smp.cpus;
1550     unsigned int max_cpus = machine->smp.max_cpus;
1551 
1552     /*
1553      * In accelerated mode, the memory map is computed earlier in kvm_type()
1554      * to create a VM with the right number of IPA bits.
1555      */
1556     if (!vms->memmap) {
1557         virt_set_memmap(vms);
1558     }
1559 
1560     /* We can probe only here because during property set
1561      * KVM is not available yet
1562      */
1563     if (vms->gic_version <= 0) {
1564         /* "host" or "max" */
1565         if (!kvm_enabled()) {
1566             if (vms->gic_version == 0) {
1567                 error_report("gic-version=host requires KVM");
1568                 exit(1);
1569             } else {
1570                 /* "max": currently means 3 for TCG */
1571                 vms->gic_version = 3;
1572             }
1573         } else {
1574             vms->gic_version = kvm_arm_vgic_probe();
1575             if (!vms->gic_version) {
1576                 error_report(
1577                     "Unable to determine GIC version supported by host");
1578                 exit(1);
1579             }
1580         }
1581     }
1582 
1583     if (!cpu_type_valid(machine->cpu_type)) {
1584         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1585         exit(1);
1586     }
1587 
1588     if (vms->secure) {
1589         if (kvm_enabled()) {
1590             error_report("mach-virt: KVM does not support Security extensions");
1591             exit(1);
1592         }
1593 
1594         /*
1595          * The Secure view of the world is the same as the NonSecure,
1596          * but with a few extra devices. Create it as a container region
1597          * containing the system memory at low priority; any secure-only
1598          * devices go in at higher priority and take precedence.
1599          */
1600         secure_sysmem = g_new(MemoryRegion, 1);
1601         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1602                            UINT64_MAX);
1603         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1604     }
1605 
1606     firmware_loaded = virt_firmware_init(vms, sysmem,
1607                                          secure_sysmem ?: sysmem);
1608 
1609     /* If we have an EL3 boot ROM then the assumption is that it will
1610      * implement PSCI itself, so disable QEMU's internal implementation
1611      * so it doesn't get in the way. Instead of starting secondary
1612      * CPUs in PSCI powerdown state we will start them all running and
1613      * let the boot ROM sort them out.
1614      * The usual case is that we do use QEMU's PSCI implementation;
1615      * if the guest has EL2 then we will use SMC as the conduit,
1616      * and otherwise we will use HVC (for backwards compatibility and
1617      * because if we're using KVM then we must use HVC).
1618      */
1619     if (vms->secure && firmware_loaded) {
1620         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1621     } else if (vms->virt) {
1622         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1623     } else {
1624         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1625     }
1626 
1627     /* The maximum number of CPUs depends on the GIC version, or on how
1628      * many redistributors we can fit into the memory map.
1629      */
1630     if (vms->gic_version == 3) {
1631         virt_max_cpus =
1632             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1633         virt_max_cpus +=
1634             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1635     } else {
1636         virt_max_cpus = GIC_NCPU;
1637     }
1638 
1639     if (max_cpus > virt_max_cpus) {
1640         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1641                      "supported by machine 'mach-virt' (%d)",
1642                      max_cpus, virt_max_cpus);
1643         exit(1);
1644     }
1645 
1646     vms->smp_cpus = smp_cpus;
1647 
1648     if (vms->virt && kvm_enabled()) {
1649         error_report("mach-virt: KVM does not support providing "
1650                      "Virtualization extensions to the guest CPU");
1651         exit(1);
1652     }
1653 
1654     create_fdt(vms);
1655 
1656     possible_cpus = mc->possible_cpu_arch_ids(machine);
1657     for (n = 0; n < possible_cpus->len; n++) {
1658         Object *cpuobj;
1659         CPUState *cs;
1660 
1661         if (n >= smp_cpus) {
1662             break;
1663         }
1664 
1665         cpuobj = object_new(possible_cpus->cpus[n].type);
1666         object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
1667                                 "mp-affinity", NULL);
1668 
1669         cs = CPU(cpuobj);
1670         cs->cpu_index = n;
1671 
1672         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1673                           &error_fatal);
1674 
1675         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1676 
1677         if (!vms->secure) {
1678             object_property_set_bool(cpuobj, false, "has_el3", NULL);
1679         }
1680 
1681         if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
1682             object_property_set_bool(cpuobj, false, "has_el2", NULL);
1683         }
1684 
1685         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1686             object_property_set_int(cpuobj, vms->psci_conduit,
1687                                     "psci-conduit", NULL);
1688 
1689             /* Secondary CPUs start in PSCI powered-down state */
1690             if (n > 0) {
1691                 object_property_set_bool(cpuobj, true,
1692                                          "start-powered-off", NULL);
1693             }
1694         }
1695 
1696         if (vmc->kvm_no_adjvtime &&
1697             object_property_find(cpuobj, "kvm-no-adjvtime", NULL)) {
1698             object_property_set_bool(cpuobj, true, "kvm-no-adjvtime", NULL);
1699         }
1700 
1701         if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1702             object_property_set_bool(cpuobj, false, "pmu", NULL);
1703         }
1704 
1705         if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1706             object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
1707                                     "reset-cbar", &error_abort);
1708         }
1709 
1710         object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1711                                  &error_abort);
1712         if (vms->secure) {
1713             object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1714                                      "secure-memory", &error_abort);
1715         }
1716 
1717         object_property_set_bool(cpuobj, true, "realized", &error_fatal);
1718         object_unref(cpuobj);
1719     }
1720     fdt_add_timer_nodes(vms);
1721     fdt_add_cpu_nodes(vms);
1722 
1723    if (!kvm_enabled()) {
1724         ARMCPU *cpu = ARM_CPU(first_cpu);
1725         bool aarch64 = object_property_get_bool(OBJECT(cpu), "aarch64", NULL);
1726 
1727         if (aarch64 && vms->highmem) {
1728             int requested_pa_size, pamax = arm_pamax(cpu);
1729 
1730             requested_pa_size = 64 - clz64(vms->highest_gpa);
1731             if (pamax < requested_pa_size) {
1732                 error_report("VCPU supports less PA bits (%d) than requested "
1733                             "by the memory map (%d)", pamax, requested_pa_size);
1734                 exit(1);
1735             }
1736         }
1737     }
1738 
1739     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
1740                                 machine->ram);
1741     if (machine->device_memory) {
1742         memory_region_add_subregion(sysmem, machine->device_memory->base,
1743                                     &machine->device_memory->mr);
1744     }
1745 
1746     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
1747 
1748     create_gic(vms);
1749 
1750     fdt_add_pmu_nodes(vms);
1751 
1752     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
1753 
1754     if (vms->secure) {
1755         create_secure_ram(vms, secure_sysmem);
1756         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
1757     }
1758 
1759     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
1760 
1761     create_rtc(vms);
1762 
1763     create_pcie(vms);
1764 
1765     if (has_ged && aarch64 && firmware_loaded && acpi_enabled) {
1766         vms->acpi_dev = create_acpi_ged(vms);
1767     } else {
1768         create_gpio(vms);
1769     }
1770 
1771      /* connect powerdown request */
1772      vms->powerdown_notifier.notify = virt_powerdown_req;
1773      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
1774 
1775     /* Create mmio transports, so the user can create virtio backends
1776      * (which will be automatically plugged in to the transports). If
1777      * no backend is created the transport will just sit harmlessly idle.
1778      */
1779     create_virtio_devices(vms);
1780 
1781     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1782     rom_set_fw(vms->fw_cfg);
1783 
1784     create_platform_bus(vms);
1785 
1786     vms->bootinfo.ram_size = machine->ram_size;
1787     vms->bootinfo.nb_cpus = smp_cpus;
1788     vms->bootinfo.board_id = -1;
1789     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1790     vms->bootinfo.get_dtb = machvirt_dtb;
1791     vms->bootinfo.skip_dtb_autoload = true;
1792     vms->bootinfo.firmware_loaded = firmware_loaded;
1793     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
1794 
1795     vms->machine_done.notify = virt_machine_done;
1796     qemu_add_machine_init_done_notifier(&vms->machine_done);
1797 }
1798 
1799 static bool virt_get_secure(Object *obj, Error **errp)
1800 {
1801     VirtMachineState *vms = VIRT_MACHINE(obj);
1802 
1803     return vms->secure;
1804 }
1805 
1806 static void virt_set_secure(Object *obj, bool value, Error **errp)
1807 {
1808     VirtMachineState *vms = VIRT_MACHINE(obj);
1809 
1810     vms->secure = value;
1811 }
1812 
1813 static bool virt_get_virt(Object *obj, Error **errp)
1814 {
1815     VirtMachineState *vms = VIRT_MACHINE(obj);
1816 
1817     return vms->virt;
1818 }
1819 
1820 static void virt_set_virt(Object *obj, bool value, Error **errp)
1821 {
1822     VirtMachineState *vms = VIRT_MACHINE(obj);
1823 
1824     vms->virt = value;
1825 }
1826 
1827 static bool virt_get_highmem(Object *obj, Error **errp)
1828 {
1829     VirtMachineState *vms = VIRT_MACHINE(obj);
1830 
1831     return vms->highmem;
1832 }
1833 
1834 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1835 {
1836     VirtMachineState *vms = VIRT_MACHINE(obj);
1837 
1838     vms->highmem = value;
1839 }
1840 
1841 static bool virt_get_its(Object *obj, Error **errp)
1842 {
1843     VirtMachineState *vms = VIRT_MACHINE(obj);
1844 
1845     return vms->its;
1846 }
1847 
1848 static void virt_set_its(Object *obj, bool value, Error **errp)
1849 {
1850     VirtMachineState *vms = VIRT_MACHINE(obj);
1851 
1852     vms->its = value;
1853 }
1854 
1855 static char *virt_get_gic_version(Object *obj, Error **errp)
1856 {
1857     VirtMachineState *vms = VIRT_MACHINE(obj);
1858     const char *val = vms->gic_version == 3 ? "3" : "2";
1859 
1860     return g_strdup(val);
1861 }
1862 
1863 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1864 {
1865     VirtMachineState *vms = VIRT_MACHINE(obj);
1866 
1867     if (!strcmp(value, "3")) {
1868         vms->gic_version = 3;
1869     } else if (!strcmp(value, "2")) {
1870         vms->gic_version = 2;
1871     } else if (!strcmp(value, "host")) {
1872         vms->gic_version = 0; /* Will probe later */
1873     } else if (!strcmp(value, "max")) {
1874         vms->gic_version = -1; /* Will probe later */
1875     } else {
1876         error_setg(errp, "Invalid gic-version value");
1877         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
1878     }
1879 }
1880 
1881 static char *virt_get_iommu(Object *obj, Error **errp)
1882 {
1883     VirtMachineState *vms = VIRT_MACHINE(obj);
1884 
1885     switch (vms->iommu) {
1886     case VIRT_IOMMU_NONE:
1887         return g_strdup("none");
1888     case VIRT_IOMMU_SMMUV3:
1889         return g_strdup("smmuv3");
1890     default:
1891         g_assert_not_reached();
1892     }
1893 }
1894 
1895 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
1896 {
1897     VirtMachineState *vms = VIRT_MACHINE(obj);
1898 
1899     if (!strcmp(value, "smmuv3")) {
1900         vms->iommu = VIRT_IOMMU_SMMUV3;
1901     } else if (!strcmp(value, "none")) {
1902         vms->iommu = VIRT_IOMMU_NONE;
1903     } else {
1904         error_setg(errp, "Invalid iommu value");
1905         error_append_hint(errp, "Valid values are none, smmuv3.\n");
1906     }
1907 }
1908 
1909 static CpuInstanceProperties
1910 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
1911 {
1912     MachineClass *mc = MACHINE_GET_CLASS(ms);
1913     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1914 
1915     assert(cpu_index < possible_cpus->len);
1916     return possible_cpus->cpus[cpu_index].props;
1917 }
1918 
1919 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
1920 {
1921     return idx % ms->numa_state->num_nodes;
1922 }
1923 
1924 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
1925 {
1926     int n;
1927     unsigned int max_cpus = ms->smp.max_cpus;
1928     VirtMachineState *vms = VIRT_MACHINE(ms);
1929 
1930     if (ms->possible_cpus) {
1931         assert(ms->possible_cpus->len == max_cpus);
1932         return ms->possible_cpus;
1933     }
1934 
1935     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1936                                   sizeof(CPUArchId) * max_cpus);
1937     ms->possible_cpus->len = max_cpus;
1938     for (n = 0; n < ms->possible_cpus->len; n++) {
1939         ms->possible_cpus->cpus[n].type = ms->cpu_type;
1940         ms->possible_cpus->cpus[n].arch_id =
1941             virt_cpu_mp_affinity(vms, n);
1942         ms->possible_cpus->cpus[n].props.has_thread_id = true;
1943         ms->possible_cpus->cpus[n].props.thread_id = n;
1944     }
1945     return ms->possible_cpus;
1946 }
1947 
1948 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1949                                  Error **errp)
1950 {
1951     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1952     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1953 
1954     if (is_nvdimm) {
1955         error_setg(errp, "nvdimm is not yet supported");
1956         return;
1957     }
1958 
1959     if (!vms->acpi_dev) {
1960         error_setg(errp,
1961                    "memory hotplug is not enabled: missing acpi-ged device");
1962         return;
1963     }
1964 
1965     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
1966 }
1967 
1968 static void virt_memory_plug(HotplugHandler *hotplug_dev,
1969                              DeviceState *dev, Error **errp)
1970 {
1971     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1972     Error *local_err = NULL;
1973 
1974     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms), &local_err);
1975     if (local_err) {
1976         goto out;
1977     }
1978 
1979     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
1980                          dev, &error_abort);
1981 
1982 out:
1983     error_propagate(errp, local_err);
1984 }
1985 
1986 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1987                                             DeviceState *dev, Error **errp)
1988 {
1989     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1990         virt_memory_pre_plug(hotplug_dev, dev, errp);
1991     }
1992 }
1993 
1994 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1995                                         DeviceState *dev, Error **errp)
1996 {
1997     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1998 
1999     if (vms->platform_bus_dev) {
2000         if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
2001             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2002                                      SYS_BUS_DEVICE(dev));
2003         }
2004     }
2005     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2006         virt_memory_plug(hotplug_dev, dev, errp);
2007     }
2008     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2009         PCIDevice *pdev = PCI_DEVICE(dev);
2010 
2011         vms->iommu = VIRT_IOMMU_VIRTIO;
2012         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2013         create_virtio_iommu_dt_bindings(vms, errp);
2014     }
2015 }
2016 
2017 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2018                                           DeviceState *dev, Error **errp)
2019 {
2020     error_setg(errp, "device unplug request for unsupported device"
2021                " type: %s", object_get_typename(OBJECT(dev)));
2022 }
2023 
2024 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2025                                                         DeviceState *dev)
2026 {
2027     if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE) ||
2028        (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
2029         return HOTPLUG_HANDLER(machine);
2030     }
2031     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2032         VirtMachineState *vms = VIRT_MACHINE(machine);
2033 
2034         if (!vms->bootinfo.firmware_loaded || !acpi_enabled) {
2035             return HOTPLUG_HANDLER(machine);
2036         }
2037     }
2038     return NULL;
2039 }
2040 
2041 /*
2042  * for arm64 kvm_type [7-0] encodes the requested number of bits
2043  * in the IPA address space
2044  */
2045 static int virt_kvm_type(MachineState *ms, const char *type_str)
2046 {
2047     VirtMachineState *vms = VIRT_MACHINE(ms);
2048     int max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms);
2049     int requested_pa_size;
2050 
2051     /* we freeze the memory map to compute the highest gpa */
2052     virt_set_memmap(vms);
2053 
2054     requested_pa_size = 64 - clz64(vms->highest_gpa);
2055 
2056     if (requested_pa_size > max_vm_pa_size) {
2057         error_report("-m and ,maxmem option values "
2058                      "require an IPA range (%d bits) larger than "
2059                      "the one supported by the host (%d bits)",
2060                      requested_pa_size, max_vm_pa_size);
2061        exit(1);
2062     }
2063     /*
2064      * By default we return 0 which corresponds to an implicit legacy
2065      * 40b IPA setting. Otherwise we return the actual requested PA
2066      * logsize
2067      */
2068     return requested_pa_size > 40 ? requested_pa_size : 0;
2069 }
2070 
2071 static void virt_machine_class_init(ObjectClass *oc, void *data)
2072 {
2073     MachineClass *mc = MACHINE_CLASS(oc);
2074     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2075 
2076     mc->init = machvirt_init;
2077     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2078      * The value may be reduced later when we have more information about the
2079      * configuration of the particular instance.
2080      */
2081     mc->max_cpus = 512;
2082     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2083     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2084     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2085     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2086     mc->block_default_type = IF_VIRTIO;
2087     mc->no_cdrom = 1;
2088     mc->pci_allow_0_address = true;
2089     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2090     mc->minimum_page_bits = 12;
2091     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2092     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2093     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2094     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2095     mc->kvm_type = virt_kvm_type;
2096     assert(!mc->get_hotplug_handler);
2097     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2098     hc->pre_plug = virt_machine_device_pre_plug_cb;
2099     hc->plug = virt_machine_device_plug_cb;
2100     hc->unplug_request = virt_machine_device_unplug_request_cb;
2101     mc->numa_mem_supported = true;
2102     mc->auto_enable_numa_with_memhp = true;
2103     mc->default_ram_id = "mach-virt.ram";
2104 }
2105 
2106 static void virt_instance_init(Object *obj)
2107 {
2108     VirtMachineState *vms = VIRT_MACHINE(obj);
2109     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2110 
2111     /* EL3 is disabled by default on virt: this makes us consistent
2112      * between KVM and TCG for this board, and it also allows us to
2113      * boot UEFI blobs which assume no TrustZone support.
2114      */
2115     vms->secure = false;
2116     object_property_add_bool(obj, "secure", virt_get_secure,
2117                              virt_set_secure, NULL);
2118     object_property_set_description(obj, "secure",
2119                                     "Set on/off to enable/disable the ARM "
2120                                     "Security Extensions (TrustZone)",
2121                                     NULL);
2122 
2123     /* EL2 is also disabled by default, for similar reasons */
2124     vms->virt = false;
2125     object_property_add_bool(obj, "virtualization", virt_get_virt,
2126                              virt_set_virt, NULL);
2127     object_property_set_description(obj, "virtualization",
2128                                     "Set on/off to enable/disable emulating a "
2129                                     "guest CPU which implements the ARM "
2130                                     "Virtualization Extensions",
2131                                     NULL);
2132 
2133     /* High memory is enabled by default */
2134     vms->highmem = true;
2135     object_property_add_bool(obj, "highmem", virt_get_highmem,
2136                              virt_set_highmem, NULL);
2137     object_property_set_description(obj, "highmem",
2138                                     "Set on/off to enable/disable using "
2139                                     "physical address space above 32 bits",
2140                                     NULL);
2141     /* Default GIC type is v2 */
2142     vms->gic_version = 2;
2143     object_property_add_str(obj, "gic-version", virt_get_gic_version,
2144                         virt_set_gic_version, NULL);
2145     object_property_set_description(obj, "gic-version",
2146                                     "Set GIC version. "
2147                                     "Valid values are 2, 3 and host", NULL);
2148 
2149     vms->highmem_ecam = !vmc->no_highmem_ecam;
2150 
2151     if (vmc->no_its) {
2152         vms->its = false;
2153     } else {
2154         /* Default allows ITS instantiation */
2155         vms->its = true;
2156         object_property_add_bool(obj, "its", virt_get_its,
2157                                  virt_set_its, NULL);
2158         object_property_set_description(obj, "its",
2159                                         "Set on/off to enable/disable "
2160                                         "ITS instantiation",
2161                                         NULL);
2162     }
2163 
2164     /* Default disallows iommu instantiation */
2165     vms->iommu = VIRT_IOMMU_NONE;
2166     object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu, NULL);
2167     object_property_set_description(obj, "iommu",
2168                                     "Set the IOMMU type. "
2169                                     "Valid values are none and smmuv3",
2170                                     NULL);
2171 
2172     vms->irqmap = a15irqmap;
2173 
2174     virt_flash_create(vms);
2175 }
2176 
2177 static const TypeInfo virt_machine_info = {
2178     .name          = TYPE_VIRT_MACHINE,
2179     .parent        = TYPE_MACHINE,
2180     .abstract      = true,
2181     .instance_size = sizeof(VirtMachineState),
2182     .class_size    = sizeof(VirtMachineClass),
2183     .class_init    = virt_machine_class_init,
2184     .instance_init = virt_instance_init,
2185     .interfaces = (InterfaceInfo[]) {
2186          { TYPE_HOTPLUG_HANDLER },
2187          { }
2188     },
2189 };
2190 
2191 static void machvirt_machine_init(void)
2192 {
2193     type_register_static(&virt_machine_info);
2194 }
2195 type_init(machvirt_machine_init);
2196 
2197 static void virt_machine_5_0_options(MachineClass *mc)
2198 {
2199 }
2200 DEFINE_VIRT_MACHINE_AS_LATEST(5, 0)
2201 
2202 static void virt_machine_4_2_options(MachineClass *mc)
2203 {
2204     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2205 
2206     virt_machine_5_0_options(mc);
2207     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
2208     vmc->kvm_no_adjvtime = true;
2209 }
2210 DEFINE_VIRT_MACHINE(4, 2)
2211 
2212 static void virt_machine_4_1_options(MachineClass *mc)
2213 {
2214     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2215 
2216     virt_machine_4_2_options(mc);
2217     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2218     vmc->no_ged = true;
2219     mc->auto_enable_numa_with_memhp = false;
2220 }
2221 DEFINE_VIRT_MACHINE(4, 1)
2222 
2223 static void virt_machine_4_0_options(MachineClass *mc)
2224 {
2225     virt_machine_4_1_options(mc);
2226     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2227 }
2228 DEFINE_VIRT_MACHINE(4, 0)
2229 
2230 static void virt_machine_3_1_options(MachineClass *mc)
2231 {
2232     virt_machine_4_0_options(mc);
2233     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2234 }
2235 DEFINE_VIRT_MACHINE(3, 1)
2236 
2237 static void virt_machine_3_0_options(MachineClass *mc)
2238 {
2239     virt_machine_3_1_options(mc);
2240     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2241 }
2242 DEFINE_VIRT_MACHINE(3, 0)
2243 
2244 static void virt_machine_2_12_options(MachineClass *mc)
2245 {
2246     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2247 
2248     virt_machine_3_0_options(mc);
2249     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2250     vmc->no_highmem_ecam = true;
2251     mc->max_cpus = 255;
2252 }
2253 DEFINE_VIRT_MACHINE(2, 12)
2254 
2255 static void virt_machine_2_11_options(MachineClass *mc)
2256 {
2257     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2258 
2259     virt_machine_2_12_options(mc);
2260     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2261     vmc->smbios_old_sys_ver = true;
2262 }
2263 DEFINE_VIRT_MACHINE(2, 11)
2264 
2265 static void virt_machine_2_10_options(MachineClass *mc)
2266 {
2267     virt_machine_2_11_options(mc);
2268     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2269     /* before 2.11 we never faulted accesses to bad addresses */
2270     mc->ignore_memory_transaction_failures = true;
2271 }
2272 DEFINE_VIRT_MACHINE(2, 10)
2273 
2274 static void virt_machine_2_9_options(MachineClass *mc)
2275 {
2276     virt_machine_2_10_options(mc);
2277     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2278 }
2279 DEFINE_VIRT_MACHINE(2, 9)
2280 
2281 static void virt_machine_2_8_options(MachineClass *mc)
2282 {
2283     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2284 
2285     virt_machine_2_9_options(mc);
2286     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
2287     /* For 2.8 and earlier we falsely claimed in the DT that
2288      * our timers were edge-triggered, not level-triggered.
2289      */
2290     vmc->claim_edge_triggered_timers = true;
2291 }
2292 DEFINE_VIRT_MACHINE(2, 8)
2293 
2294 static void virt_machine_2_7_options(MachineClass *mc)
2295 {
2296     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2297 
2298     virt_machine_2_8_options(mc);
2299     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2300     /* ITS was introduced with 2.8 */
2301     vmc->no_its = true;
2302     /* Stick with 1K pages for migration compatibility */
2303     mc->minimum_page_bits = 0;
2304 }
2305 DEFINE_VIRT_MACHINE(2, 7)
2306 
2307 static void virt_machine_2_6_options(MachineClass *mc)
2308 {
2309     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2310 
2311     virt_machine_2_7_options(mc);
2312     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
2313     vmc->disallow_affinity_adjustment = true;
2314     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2315     vmc->no_pmu = true;
2316 }
2317 DEFINE_VIRT_MACHINE(2, 6)
2318