xref: /openbmc/qemu/hw/arm/virt.c (revision 40bd0502)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu-common.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "qapi/error.h"
37 #include "hw/sysbus.h"
38 #include "hw/boards.h"
39 #include "hw/arm/boot.h"
40 #include "hw/arm/primecell.h"
41 #include "hw/arm/virt.h"
42 #include "hw/block/flash.h"
43 #include "hw/vfio/vfio-calxeda-xgmac.h"
44 #include "hw/vfio/vfio-amd-xgbe.h"
45 #include "hw/display/ramfb.h"
46 #include "net/net.h"
47 #include "sysemu/device_tree.h"
48 #include "sysemu/numa.h"
49 #include "sysemu/runstate.h"
50 #include "sysemu/sysemu.h"
51 #include "sysemu/tpm.h"
52 #include "sysemu/kvm.h"
53 #include "hw/loader.h"
54 #include "exec/address-spaces.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/arm/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/irq.h"
67 #include "kvm_arm.h"
68 #include "hw/firmware/smbios.h"
69 #include "qapi/visitor.h"
70 #include "qapi/qapi-visit-common.h"
71 #include "standard-headers/linux/input.h"
72 #include "hw/arm/smmuv3.h"
73 #include "hw/acpi/acpi.h"
74 #include "target/arm/internals.h"
75 #include "hw/mem/pc-dimm.h"
76 #include "hw/mem/nvdimm.h"
77 #include "hw/acpi/generic_event_device.h"
78 #include "hw/virtio/virtio-iommu.h"
79 #include "hw/char/pl011.h"
80 
81 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
82     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
83                                                     void *data) \
84     { \
85         MachineClass *mc = MACHINE_CLASS(oc); \
86         virt_machine_##major##_##minor##_options(mc); \
87         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
88         if (latest) { \
89             mc->alias = "virt"; \
90         } \
91     } \
92     static const TypeInfo machvirt_##major##_##minor##_info = { \
93         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
94         .parent = TYPE_VIRT_MACHINE, \
95         .class_init = virt_##major##_##minor##_class_init, \
96     }; \
97     static void machvirt_machine_##major##_##minor##_init(void) \
98     { \
99         type_register_static(&machvirt_##major##_##minor##_info); \
100     } \
101     type_init(machvirt_machine_##major##_##minor##_init);
102 
103 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
104     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
105 #define DEFINE_VIRT_MACHINE(major, minor) \
106     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
107 
108 
109 /* Number of external interrupt lines to configure the GIC with */
110 #define NUM_IRQS 256
111 
112 #define PLATFORM_BUS_NUM_IRQS 64
113 
114 /* Legacy RAM limit in GB (< version 4.0) */
115 #define LEGACY_RAMLIMIT_GB 255
116 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
117 
118 /* Addresses and sizes of our components.
119  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
120  * 128MB..256MB is used for miscellaneous device I/O.
121  * 256MB..1GB is reserved for possible future PCI support (ie where the
122  * PCI memory window will go if we add a PCI host controller).
123  * 1GB and up is RAM (which may happily spill over into the
124  * high memory region beyond 4GB).
125  * This represents a compromise between how much RAM can be given to
126  * a 32 bit VM and leaving space for expansion and in particular for PCI.
127  * Note that devices should generally be placed at multiples of 0x10000,
128  * to accommodate guests using 64K pages.
129  */
130 static const MemMapEntry base_memmap[] = {
131     /* Space up to 0x8000000 is reserved for a boot ROM */
132     [VIRT_FLASH] =              {          0, 0x08000000 },
133     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
134     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
135     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
136     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
137     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
138     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
139     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
140     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
141     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
142     /* This redistributor space allows up to 2*64kB*123 CPUs */
143     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
144     [VIRT_UART] =               { 0x09000000, 0x00001000 },
145     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
146     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
147     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
148     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
149     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
150     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
151     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
152     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
153     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
154     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
155     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
156     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
157     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
158     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
159     /* Actual RAM size depends on initial RAM and device memory settings */
160     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
161 };
162 
163 /*
164  * Highmem IO Regions: This memory map is floating, located after the RAM.
165  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
166  * top of the RAM, so that its base get the same alignment as the size,
167  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
168  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
169  * Note the extended_memmap is sized so that it eventually also includes the
170  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
171  * index of base_memmap).
172  */
173 static MemMapEntry extended_memmap[] = {
174     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
175     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
176     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
177     /* Second PCIe window */
178     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
179 };
180 
181 static const int a15irqmap[] = {
182     [VIRT_UART] = 1,
183     [VIRT_RTC] = 2,
184     [VIRT_PCIE] = 3, /* ... to 6 */
185     [VIRT_GPIO] = 7,
186     [VIRT_SECURE_UART] = 8,
187     [VIRT_ACPI_GED] = 9,
188     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
189     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
190     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
191     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
192 };
193 
194 static const char *valid_cpus[] = {
195     ARM_CPU_TYPE_NAME("cortex-a7"),
196     ARM_CPU_TYPE_NAME("cortex-a15"),
197     ARM_CPU_TYPE_NAME("cortex-a53"),
198     ARM_CPU_TYPE_NAME("cortex-a57"),
199     ARM_CPU_TYPE_NAME("cortex-a72"),
200     ARM_CPU_TYPE_NAME("host"),
201     ARM_CPU_TYPE_NAME("max"),
202 };
203 
204 static bool cpu_type_valid(const char *cpu)
205 {
206     int i;
207 
208     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
209         if (strcmp(cpu, valid_cpus[i]) == 0) {
210             return true;
211         }
212     }
213     return false;
214 }
215 
216 static void create_fdt(VirtMachineState *vms)
217 {
218     MachineState *ms = MACHINE(vms);
219     int nb_numa_nodes = ms->numa_state->num_nodes;
220     void *fdt = create_device_tree(&vms->fdt_size);
221 
222     if (!fdt) {
223         error_report("create_device_tree() failed");
224         exit(1);
225     }
226 
227     vms->fdt = fdt;
228 
229     /* Header */
230     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
231     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
232     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
233 
234     /* /chosen must exist for load_dtb to fill in necessary properties later */
235     qemu_fdt_add_subnode(fdt, "/chosen");
236 
237     /* Clock node, for the benefit of the UART. The kernel device tree
238      * binding documentation claims the PL011 node clock properties are
239      * optional but in practice if you omit them the kernel refuses to
240      * probe for the device.
241      */
242     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
243     qemu_fdt_add_subnode(fdt, "/apb-pclk");
244     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
245     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
246     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
247     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
248                                 "clk24mhz");
249     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
250 
251     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
252         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
253         uint32_t *matrix = g_malloc0(size);
254         int idx, i, j;
255 
256         for (i = 0; i < nb_numa_nodes; i++) {
257             for (j = 0; j < nb_numa_nodes; j++) {
258                 idx = (i * nb_numa_nodes + j) * 3;
259                 matrix[idx + 0] = cpu_to_be32(i);
260                 matrix[idx + 1] = cpu_to_be32(j);
261                 matrix[idx + 2] =
262                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
263             }
264         }
265 
266         qemu_fdt_add_subnode(fdt, "/distance-map");
267         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
268                                 "numa-distance-map-v1");
269         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
270                          matrix, size);
271         g_free(matrix);
272     }
273 }
274 
275 static void fdt_add_timer_nodes(const VirtMachineState *vms)
276 {
277     /* On real hardware these interrupts are level-triggered.
278      * On KVM they were edge-triggered before host kernel version 4.4,
279      * and level-triggered afterwards.
280      * On emulated QEMU they are level-triggered.
281      *
282      * Getting the DTB info about them wrong is awkward for some
283      * guest kernels:
284      *  pre-4.8 ignore the DT and leave the interrupt configured
285      *   with whatever the GIC reset value (or the bootloader) left it at
286      *  4.8 before rc6 honour the incorrect data by programming it back
287      *   into the GIC, causing problems
288      *  4.8rc6 and later ignore the DT and always write "level triggered"
289      *   into the GIC
290      *
291      * For backwards-compatibility, virt-2.8 and earlier will continue
292      * to say these are edge-triggered, but later machines will report
293      * the correct information.
294      */
295     ARMCPU *armcpu;
296     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
297     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
298 
299     if (vmc->claim_edge_triggered_timers) {
300         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
301     }
302 
303     if (vms->gic_version == VIRT_GIC_VERSION_2) {
304         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
305                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
306                              (1 << vms->smp_cpus) - 1);
307     }
308 
309     qemu_fdt_add_subnode(vms->fdt, "/timer");
310 
311     armcpu = ARM_CPU(qemu_get_cpu(0));
312     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
313         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
314         qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
315                          compat, sizeof(compat));
316     } else {
317         qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
318                                 "arm,armv7-timer");
319     }
320     qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
321     qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
322                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
323                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
324                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
325                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
326 }
327 
328 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
329 {
330     int cpu;
331     int addr_cells = 1;
332     const MachineState *ms = MACHINE(vms);
333 
334     /*
335      * From Documentation/devicetree/bindings/arm/cpus.txt
336      *  On ARM v8 64-bit systems value should be set to 2,
337      *  that corresponds to the MPIDR_EL1 register size.
338      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
339      *  in the system, #address-cells can be set to 1, since
340      *  MPIDR_EL1[63:32] bits are not used for CPUs
341      *  identification.
342      *
343      *  Here we actually don't know whether our system is 32- or 64-bit one.
344      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
345      *  at least one of them has Aff3 populated, we set #address-cells to 2.
346      */
347     for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
348         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
349 
350         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
351             addr_cells = 2;
352             break;
353         }
354     }
355 
356     qemu_fdt_add_subnode(vms->fdt, "/cpus");
357     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
358     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
359 
360     for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
361         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
362         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
363         CPUState *cs = CPU(armcpu);
364 
365         qemu_fdt_add_subnode(vms->fdt, nodename);
366         qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
367         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
368                                     armcpu->dtb_compatible);
369 
370         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
371             && vms->smp_cpus > 1) {
372             qemu_fdt_setprop_string(vms->fdt, nodename,
373                                         "enable-method", "psci");
374         }
375 
376         if (addr_cells == 2) {
377             qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
378                                  armcpu->mp_affinity);
379         } else {
380             qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
381                                   armcpu->mp_affinity);
382         }
383 
384         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
385             qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
386                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
387         }
388 
389         g_free(nodename);
390     }
391 }
392 
393 static void fdt_add_its_gic_node(VirtMachineState *vms)
394 {
395     char *nodename;
396 
397     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
398     nodename = g_strdup_printf("/intc/its@%" PRIx64,
399                                vms->memmap[VIRT_GIC_ITS].base);
400     qemu_fdt_add_subnode(vms->fdt, nodename);
401     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
402                             "arm,gic-v3-its");
403     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
404     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
405                                  2, vms->memmap[VIRT_GIC_ITS].base,
406                                  2, vms->memmap[VIRT_GIC_ITS].size);
407     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
408     g_free(nodename);
409 }
410 
411 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
412 {
413     char *nodename;
414 
415     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
416                                vms->memmap[VIRT_GIC_V2M].base);
417     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
418     qemu_fdt_add_subnode(vms->fdt, nodename);
419     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
420                             "arm,gic-v2m-frame");
421     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
422     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
423                                  2, vms->memmap[VIRT_GIC_V2M].base,
424                                  2, vms->memmap[VIRT_GIC_V2M].size);
425     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
426     g_free(nodename);
427 }
428 
429 static void fdt_add_gic_node(VirtMachineState *vms)
430 {
431     char *nodename;
432 
433     vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
434     qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
435 
436     nodename = g_strdup_printf("/intc@%" PRIx64,
437                                vms->memmap[VIRT_GIC_DIST].base);
438     qemu_fdt_add_subnode(vms->fdt, nodename);
439     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 3);
440     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-controller", NULL, 0);
441     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 0x2);
442     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 0x2);
443     qemu_fdt_setprop(vms->fdt, nodename, "ranges", NULL, 0);
444     if (vms->gic_version == VIRT_GIC_VERSION_3) {
445         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
446 
447         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
448                                 "arm,gic-v3");
449 
450         qemu_fdt_setprop_cell(vms->fdt, nodename,
451                               "#redistributor-regions", nb_redist_regions);
452 
453         if (nb_redist_regions == 1) {
454             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
455                                          2, vms->memmap[VIRT_GIC_DIST].base,
456                                          2, vms->memmap[VIRT_GIC_DIST].size,
457                                          2, vms->memmap[VIRT_GIC_REDIST].base,
458                                          2, vms->memmap[VIRT_GIC_REDIST].size);
459         } else {
460             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
461                                  2, vms->memmap[VIRT_GIC_DIST].base,
462                                  2, vms->memmap[VIRT_GIC_DIST].size,
463                                  2, vms->memmap[VIRT_GIC_REDIST].base,
464                                  2, vms->memmap[VIRT_GIC_REDIST].size,
465                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
466                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
467         }
468 
469         if (vms->virt) {
470             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
471                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
472                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
473         }
474     } else {
475         /* 'cortex-a15-gic' means 'GIC v2' */
476         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
477                                 "arm,cortex-a15-gic");
478         if (!vms->virt) {
479             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
480                                          2, vms->memmap[VIRT_GIC_DIST].base,
481                                          2, vms->memmap[VIRT_GIC_DIST].size,
482                                          2, vms->memmap[VIRT_GIC_CPU].base,
483                                          2, vms->memmap[VIRT_GIC_CPU].size);
484         } else {
485             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
486                                          2, vms->memmap[VIRT_GIC_DIST].base,
487                                          2, vms->memmap[VIRT_GIC_DIST].size,
488                                          2, vms->memmap[VIRT_GIC_CPU].base,
489                                          2, vms->memmap[VIRT_GIC_CPU].size,
490                                          2, vms->memmap[VIRT_GIC_HYP].base,
491                                          2, vms->memmap[VIRT_GIC_HYP].size,
492                                          2, vms->memmap[VIRT_GIC_VCPU].base,
493                                          2, vms->memmap[VIRT_GIC_VCPU].size);
494             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
495                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
496                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
497         }
498     }
499 
500     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->gic_phandle);
501     g_free(nodename);
502 }
503 
504 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
505 {
506     CPUState *cpu;
507     ARMCPU *armcpu;
508     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
509 
510     CPU_FOREACH(cpu) {
511         armcpu = ARM_CPU(cpu);
512         if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
513             return;
514         }
515         if (kvm_enabled()) {
516             if (kvm_irqchip_in_kernel()) {
517                 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
518             }
519             kvm_arm_pmu_init(cpu);
520         }
521     }
522 
523     if (vms->gic_version == VIRT_GIC_VERSION_2) {
524         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
525                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
526                              (1 << vms->smp_cpus) - 1);
527     }
528 
529     armcpu = ARM_CPU(qemu_get_cpu(0));
530     qemu_fdt_add_subnode(vms->fdt, "/pmu");
531     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
532         const char compat[] = "arm,armv8-pmuv3";
533         qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
534                          compat, sizeof(compat));
535         qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
536                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
537     }
538 }
539 
540 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
541 {
542     DeviceState *dev;
543     MachineState *ms = MACHINE(vms);
544     int irq = vms->irqmap[VIRT_ACPI_GED];
545     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
546 
547     if (ms->ram_slots) {
548         event |= ACPI_GED_MEM_HOTPLUG_EVT;
549     }
550 
551     dev = qdev_create(NULL, TYPE_ACPI_GED);
552     qdev_prop_set_uint32(dev, "ged-event", event);
553 
554     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
555     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
556     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
557 
558     qdev_init_nofail(dev);
559 
560     return dev;
561 }
562 
563 static void create_its(VirtMachineState *vms)
564 {
565     const char *itsclass = its_class_name();
566     DeviceState *dev;
567 
568     if (!itsclass) {
569         /* Do nothing if not supported */
570         return;
571     }
572 
573     dev = qdev_create(NULL, itsclass);
574 
575     object_property_set_link(OBJECT(dev), OBJECT(vms->gic), "parent-gicv3",
576                              &error_abort);
577     qdev_init_nofail(dev);
578     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
579 
580     fdt_add_its_gic_node(vms);
581 }
582 
583 static void create_v2m(VirtMachineState *vms)
584 {
585     int i;
586     int irq = vms->irqmap[VIRT_GIC_V2M];
587     DeviceState *dev;
588 
589     dev = qdev_create(NULL, "arm-gicv2m");
590     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
591     qdev_prop_set_uint32(dev, "base-spi", irq);
592     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
593     qdev_init_nofail(dev);
594 
595     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
596         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
597                            qdev_get_gpio_in(vms->gic, irq + i));
598     }
599 
600     fdt_add_v2m_gic_node(vms);
601 }
602 
603 static void create_gic(VirtMachineState *vms)
604 {
605     MachineState *ms = MACHINE(vms);
606     /* We create a standalone GIC */
607     SysBusDevice *gicbusdev;
608     const char *gictype;
609     int type = vms->gic_version, i;
610     unsigned int smp_cpus = ms->smp.cpus;
611     uint32_t nb_redist_regions = 0;
612 
613     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
614 
615     vms->gic = qdev_create(NULL, gictype);
616     qdev_prop_set_uint32(vms->gic, "revision", type);
617     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
618     /* Note that the num-irq property counts both internal and external
619      * interrupts; there are always 32 of the former (mandated by GIC spec).
620      */
621     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
622     if (!kvm_irqchip_in_kernel()) {
623         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
624     }
625 
626     if (type == 3) {
627         uint32_t redist0_capacity =
628                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
629         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
630 
631         nb_redist_regions = virt_gicv3_redist_region_count(vms);
632 
633         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
634                              nb_redist_regions);
635         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
636 
637         if (nb_redist_regions == 2) {
638             uint32_t redist1_capacity =
639                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
640 
641             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
642                 MIN(smp_cpus - redist0_count, redist1_capacity));
643         }
644     } else {
645         if (!kvm_irqchip_in_kernel()) {
646             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
647                               vms->virt);
648         }
649     }
650     qdev_init_nofail(vms->gic);
651     gicbusdev = SYS_BUS_DEVICE(vms->gic);
652     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
653     if (type == 3) {
654         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
655         if (nb_redist_regions == 2) {
656             sysbus_mmio_map(gicbusdev, 2,
657                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
658         }
659     } else {
660         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
661         if (vms->virt) {
662             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
663             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
664         }
665     }
666 
667     /* Wire the outputs from each CPU's generic timer and the GICv3
668      * maintenance interrupt signal to the appropriate GIC PPI inputs,
669      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
670      */
671     for (i = 0; i < smp_cpus; i++) {
672         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
673         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
674         int irq;
675         /* Mapping from the output timer irq lines from the CPU to the
676          * GIC PPI inputs we use for the virt board.
677          */
678         const int timer_irq[] = {
679             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
680             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
681             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
682             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
683         };
684 
685         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
686             qdev_connect_gpio_out(cpudev, irq,
687                                   qdev_get_gpio_in(vms->gic,
688                                                    ppibase + timer_irq[irq]));
689         }
690 
691         if (type == 3) {
692             qemu_irq irq = qdev_get_gpio_in(vms->gic,
693                                             ppibase + ARCH_GIC_MAINT_IRQ);
694             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
695                                         0, irq);
696         } else if (vms->virt) {
697             qemu_irq irq = qdev_get_gpio_in(vms->gic,
698                                             ppibase + ARCH_GIC_MAINT_IRQ);
699             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
700         }
701 
702         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
703                                     qdev_get_gpio_in(vms->gic, ppibase
704                                                      + VIRTUAL_PMU_IRQ));
705 
706         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
707         sysbus_connect_irq(gicbusdev, i + smp_cpus,
708                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
709         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
710                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
711         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
712                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
713     }
714 
715     fdt_add_gic_node(vms);
716 
717     if (type == 3 && vms->its) {
718         create_its(vms);
719     } else if (type == 2) {
720         create_v2m(vms);
721     }
722 }
723 
724 static void create_uart(const VirtMachineState *vms, int uart,
725                         MemoryRegion *mem, Chardev *chr)
726 {
727     char *nodename;
728     hwaddr base = vms->memmap[uart].base;
729     hwaddr size = vms->memmap[uart].size;
730     int irq = vms->irqmap[uart];
731     const char compat[] = "arm,pl011\0arm,primecell";
732     const char clocknames[] = "uartclk\0apb_pclk";
733     DeviceState *dev = qdev_create(NULL, TYPE_PL011);
734     SysBusDevice *s = SYS_BUS_DEVICE(dev);
735 
736     qdev_prop_set_chr(dev, "chardev", chr);
737     qdev_init_nofail(dev);
738     memory_region_add_subregion(mem, base,
739                                 sysbus_mmio_get_region(s, 0));
740     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
741 
742     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
743     qemu_fdt_add_subnode(vms->fdt, nodename);
744     /* Note that we can't use setprop_string because of the embedded NUL */
745     qemu_fdt_setprop(vms->fdt, nodename, "compatible",
746                          compat, sizeof(compat));
747     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
748                                      2, base, 2, size);
749     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
750                                GIC_FDT_IRQ_TYPE_SPI, irq,
751                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
752     qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
753                                vms->clock_phandle, vms->clock_phandle);
754     qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
755                          clocknames, sizeof(clocknames));
756 
757     if (uart == VIRT_UART) {
758         qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
759     } else {
760         /* Mark as not usable by the normal world */
761         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
762         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
763 
764         qemu_fdt_add_subnode(vms->fdt, "/secure-chosen");
765         qemu_fdt_setprop_string(vms->fdt, "/secure-chosen", "stdout-path",
766                                 nodename);
767     }
768 
769     g_free(nodename);
770 }
771 
772 static void create_rtc(const VirtMachineState *vms)
773 {
774     char *nodename;
775     hwaddr base = vms->memmap[VIRT_RTC].base;
776     hwaddr size = vms->memmap[VIRT_RTC].size;
777     int irq = vms->irqmap[VIRT_RTC];
778     const char compat[] = "arm,pl031\0arm,primecell";
779 
780     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
781 
782     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
783     qemu_fdt_add_subnode(vms->fdt, nodename);
784     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
785     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
786                                  2, base, 2, size);
787     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
788                            GIC_FDT_IRQ_TYPE_SPI, irq,
789                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
790     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
791     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
792     g_free(nodename);
793 }
794 
795 static DeviceState *gpio_key_dev;
796 static void virt_powerdown_req(Notifier *n, void *opaque)
797 {
798     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
799 
800     if (s->acpi_dev) {
801         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
802     } else {
803         /* use gpio Pin 3 for power button event */
804         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
805     }
806 }
807 
808 static void create_gpio(const VirtMachineState *vms)
809 {
810     char *nodename;
811     DeviceState *pl061_dev;
812     hwaddr base = vms->memmap[VIRT_GPIO].base;
813     hwaddr size = vms->memmap[VIRT_GPIO].size;
814     int irq = vms->irqmap[VIRT_GPIO];
815     const char compat[] = "arm,pl061\0arm,primecell";
816 
817     pl061_dev = sysbus_create_simple("pl061", base,
818                                      qdev_get_gpio_in(vms->gic, irq));
819 
820     uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
821     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
822     qemu_fdt_add_subnode(vms->fdt, nodename);
823     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
824                                  2, base, 2, size);
825     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
826     qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
827     qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
828     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
829                            GIC_FDT_IRQ_TYPE_SPI, irq,
830                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
831     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
832     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
833     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
834 
835     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
836                                         qdev_get_gpio_in(pl061_dev, 3));
837     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
838     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
839     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
840     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
841 
842     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
843     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
844                             "label", "GPIO Key Poweroff");
845     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
846                           KEY_POWER);
847     qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
848                            "gpios", phandle, 3, 0);
849     g_free(nodename);
850 }
851 
852 static void create_virtio_devices(const VirtMachineState *vms)
853 {
854     int i;
855     hwaddr size = vms->memmap[VIRT_MMIO].size;
856 
857     /* We create the transports in forwards order. Since qbus_realize()
858      * prepends (not appends) new child buses, the incrementing loop below will
859      * create a list of virtio-mmio buses with decreasing base addresses.
860      *
861      * When a -device option is processed from the command line,
862      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
863      * order. The upshot is that -device options in increasing command line
864      * order are mapped to virtio-mmio buses with decreasing base addresses.
865      *
866      * When this code was originally written, that arrangement ensured that the
867      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
868      * the first -device on the command line. (The end-to-end order is a
869      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
870      * guest kernel's name-to-address assignment strategy.)
871      *
872      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
873      * the message, if not necessarily the code, of commit 70161ff336.
874      * Therefore the loop now establishes the inverse of the original intent.
875      *
876      * Unfortunately, we can't counteract the kernel change by reversing the
877      * loop; it would break existing command lines.
878      *
879      * In any case, the kernel makes no guarantee about the stability of
880      * enumeration order of virtio devices (as demonstrated by it changing
881      * between kernel versions). For reliable and stable identification
882      * of disks users must use UUIDs or similar mechanisms.
883      */
884     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
885         int irq = vms->irqmap[VIRT_MMIO] + i;
886         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
887 
888         sysbus_create_simple("virtio-mmio", base,
889                              qdev_get_gpio_in(vms->gic, irq));
890     }
891 
892     /* We add dtb nodes in reverse order so that they appear in the finished
893      * device tree lowest address first.
894      *
895      * Note that this mapping is independent of the loop above. The previous
896      * loop influences virtio device to virtio transport assignment, whereas
897      * this loop controls how virtio transports are laid out in the dtb.
898      */
899     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
900         char *nodename;
901         int irq = vms->irqmap[VIRT_MMIO] + i;
902         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
903 
904         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
905         qemu_fdt_add_subnode(vms->fdt, nodename);
906         qemu_fdt_setprop_string(vms->fdt, nodename,
907                                 "compatible", "virtio,mmio");
908         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
909                                      2, base, 2, size);
910         qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
911                                GIC_FDT_IRQ_TYPE_SPI, irq,
912                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
913         qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
914         g_free(nodename);
915     }
916 }
917 
918 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
919 
920 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
921                                         const char *name,
922                                         const char *alias_prop_name)
923 {
924     /*
925      * Create a single flash device.  We use the same parameters as
926      * the flash devices on the Versatile Express board.
927      */
928     DeviceState *dev = qdev_create(NULL, TYPE_PFLASH_CFI01);
929 
930     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
931     qdev_prop_set_uint8(dev, "width", 4);
932     qdev_prop_set_uint8(dev, "device-width", 2);
933     qdev_prop_set_bit(dev, "big-endian", false);
934     qdev_prop_set_uint16(dev, "id0", 0x89);
935     qdev_prop_set_uint16(dev, "id1", 0x18);
936     qdev_prop_set_uint16(dev, "id2", 0x00);
937     qdev_prop_set_uint16(dev, "id3", 0x00);
938     qdev_prop_set_string(dev, "name", name);
939     object_property_add_child(OBJECT(vms), name, OBJECT(dev),
940                               &error_abort);
941     object_property_add_alias(OBJECT(vms), alias_prop_name,
942                               OBJECT(dev), "drive", &error_abort);
943     return PFLASH_CFI01(dev);
944 }
945 
946 static void virt_flash_create(VirtMachineState *vms)
947 {
948     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
949     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
950 }
951 
952 static void virt_flash_map1(PFlashCFI01 *flash,
953                             hwaddr base, hwaddr size,
954                             MemoryRegion *sysmem)
955 {
956     DeviceState *dev = DEVICE(flash);
957 
958     assert(size % VIRT_FLASH_SECTOR_SIZE == 0);
959     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
960     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
961     qdev_init_nofail(dev);
962 
963     memory_region_add_subregion(sysmem, base,
964                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
965                                                        0));
966 }
967 
968 static void virt_flash_map(VirtMachineState *vms,
969                            MemoryRegion *sysmem,
970                            MemoryRegion *secure_sysmem)
971 {
972     /*
973      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
974      * sysmem is the system memory space. secure_sysmem is the secure view
975      * of the system, and the first flash device should be made visible only
976      * there. The second flash device is visible to both secure and nonsecure.
977      * If sysmem == secure_sysmem this means there is no separate Secure
978      * address space and both flash devices are generally visible.
979      */
980     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
981     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
982 
983     virt_flash_map1(vms->flash[0], flashbase, flashsize,
984                     secure_sysmem);
985     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
986                     sysmem);
987 }
988 
989 static void virt_flash_fdt(VirtMachineState *vms,
990                            MemoryRegion *sysmem,
991                            MemoryRegion *secure_sysmem)
992 {
993     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
994     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
995     char *nodename;
996 
997     if (sysmem == secure_sysmem) {
998         /* Report both flash devices as a single node in the DT */
999         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1000         qemu_fdt_add_subnode(vms->fdt, nodename);
1001         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1002         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1003                                      2, flashbase, 2, flashsize,
1004                                      2, flashbase + flashsize, 2, flashsize);
1005         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1006         g_free(nodename);
1007     } else {
1008         /*
1009          * Report the devices as separate nodes so we can mark one as
1010          * only visible to the secure world.
1011          */
1012         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1013         qemu_fdt_add_subnode(vms->fdt, nodename);
1014         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1015         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1016                                      2, flashbase, 2, flashsize);
1017         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1018         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1019         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1020         g_free(nodename);
1021 
1022         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1023         qemu_fdt_add_subnode(vms->fdt, nodename);
1024         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1025         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1026                                      2, flashbase + flashsize, 2, flashsize);
1027         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1028         g_free(nodename);
1029     }
1030 }
1031 
1032 static bool virt_firmware_init(VirtMachineState *vms,
1033                                MemoryRegion *sysmem,
1034                                MemoryRegion *secure_sysmem)
1035 {
1036     int i;
1037     BlockBackend *pflash_blk0;
1038 
1039     /* Map legacy -drive if=pflash to machine properties */
1040     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1041         pflash_cfi01_legacy_drive(vms->flash[i],
1042                                   drive_get(IF_PFLASH, 0, i));
1043     }
1044 
1045     virt_flash_map(vms, sysmem, secure_sysmem);
1046 
1047     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1048 
1049     if (bios_name) {
1050         char *fname;
1051         MemoryRegion *mr;
1052         int image_size;
1053 
1054         if (pflash_blk0) {
1055             error_report("The contents of the first flash device may be "
1056                          "specified with -bios or with -drive if=pflash... "
1057                          "but you cannot use both options at once");
1058             exit(1);
1059         }
1060 
1061         /* Fall back to -bios */
1062 
1063         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1064         if (!fname) {
1065             error_report("Could not find ROM image '%s'", bios_name);
1066             exit(1);
1067         }
1068         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1069         image_size = load_image_mr(fname, mr);
1070         g_free(fname);
1071         if (image_size < 0) {
1072             error_report("Could not load ROM image '%s'", bios_name);
1073             exit(1);
1074         }
1075     }
1076 
1077     return pflash_blk0 || bios_name;
1078 }
1079 
1080 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1081 {
1082     MachineState *ms = MACHINE(vms);
1083     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1084     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1085     FWCfgState *fw_cfg;
1086     char *nodename;
1087 
1088     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1089     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1090 
1091     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1092     qemu_fdt_add_subnode(vms->fdt, nodename);
1093     qemu_fdt_setprop_string(vms->fdt, nodename,
1094                             "compatible", "qemu,fw-cfg-mmio");
1095     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1096                                  2, base, 2, size);
1097     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1098     g_free(nodename);
1099     return fw_cfg;
1100 }
1101 
1102 static void create_pcie_irq_map(const VirtMachineState *vms,
1103                                 uint32_t gic_phandle,
1104                                 int first_irq, const char *nodename)
1105 {
1106     int devfn, pin;
1107     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1108     uint32_t *irq_map = full_irq_map;
1109 
1110     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1111         for (pin = 0; pin < 4; pin++) {
1112             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1113             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1114             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1115             int i;
1116 
1117             uint32_t map[] = {
1118                 devfn << 8, 0, 0,                           /* devfn */
1119                 pin + 1,                                    /* PCI pin */
1120                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1121 
1122             /* Convert map to big endian */
1123             for (i = 0; i < 10; i++) {
1124                 irq_map[i] = cpu_to_be32(map[i]);
1125             }
1126             irq_map += 10;
1127         }
1128     }
1129 
1130     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
1131                      full_irq_map, sizeof(full_irq_map));
1132 
1133     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
1134                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
1135                            0x7           /* PCI irq */);
1136 }
1137 
1138 static void create_smmu(const VirtMachineState *vms,
1139                         PCIBus *bus)
1140 {
1141     char *node;
1142     const char compat[] = "arm,smmu-v3";
1143     int irq =  vms->irqmap[VIRT_SMMU];
1144     int i;
1145     hwaddr base = vms->memmap[VIRT_SMMU].base;
1146     hwaddr size = vms->memmap[VIRT_SMMU].size;
1147     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1148     DeviceState *dev;
1149 
1150     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1151         return;
1152     }
1153 
1154     dev = qdev_create(NULL, "arm-smmuv3");
1155 
1156     object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
1157                              &error_abort);
1158     qdev_init_nofail(dev);
1159     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1160     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1161         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1162                            qdev_get_gpio_in(vms->gic, irq + i));
1163     }
1164 
1165     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1166     qemu_fdt_add_subnode(vms->fdt, node);
1167     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1168     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
1169 
1170     qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
1171             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1172             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1173             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1174             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1175 
1176     qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
1177                      sizeof(irq_names));
1178 
1179     qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
1180     qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
1181     qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
1182 
1183     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1184 
1185     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1186     g_free(node);
1187 }
1188 
1189 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms, Error **errp)
1190 {
1191     const char compat[] = "virtio,pci-iommu";
1192     uint16_t bdf = vms->virtio_iommu_bdf;
1193     char *node;
1194 
1195     vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1196 
1197     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1198     qemu_fdt_add_subnode(vms->fdt, node);
1199     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1200     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg",
1201                                  1, bdf << 8, 1, 0, 1, 0,
1202                                  1, 0, 1, 0);
1203 
1204     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1205     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1206     g_free(node);
1207 
1208     qemu_fdt_setprop_cells(vms->fdt, vms->pciehb_nodename, "iommu-map",
1209                            0x0, vms->iommu_phandle, 0x0, bdf,
1210                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1211 }
1212 
1213 static void create_pcie(VirtMachineState *vms)
1214 {
1215     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1216     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1217     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1218     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1219     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1220     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1221     hwaddr base_ecam, size_ecam;
1222     hwaddr base = base_mmio;
1223     int nr_pcie_buses;
1224     int irq = vms->irqmap[VIRT_PCIE];
1225     MemoryRegion *mmio_alias;
1226     MemoryRegion *mmio_reg;
1227     MemoryRegion *ecam_alias;
1228     MemoryRegion *ecam_reg;
1229     DeviceState *dev;
1230     char *nodename;
1231     int i, ecam_id;
1232     PCIHostState *pci;
1233 
1234     dev = qdev_create(NULL, TYPE_GPEX_HOST);
1235     qdev_init_nofail(dev);
1236 
1237     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1238     base_ecam = vms->memmap[ecam_id].base;
1239     size_ecam = vms->memmap[ecam_id].size;
1240     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1241     /* Map only the first size_ecam bytes of ECAM space */
1242     ecam_alias = g_new0(MemoryRegion, 1);
1243     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1244     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1245                              ecam_reg, 0, size_ecam);
1246     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1247 
1248     /* Map the MMIO window into system address space so as to expose
1249      * the section of PCI MMIO space which starts at the same base address
1250      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1251      * the window).
1252      */
1253     mmio_alias = g_new0(MemoryRegion, 1);
1254     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1255     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1256                              mmio_reg, base_mmio, size_mmio);
1257     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1258 
1259     if (vms->highmem) {
1260         /* Map high MMIO space */
1261         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1262 
1263         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1264                                  mmio_reg, base_mmio_high, size_mmio_high);
1265         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1266                                     high_mmio_alias);
1267     }
1268 
1269     /* Map IO port space */
1270     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1271 
1272     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1273         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1274                            qdev_get_gpio_in(vms->gic, irq + i));
1275         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1276     }
1277 
1278     pci = PCI_HOST_BRIDGE(dev);
1279     if (pci->bus) {
1280         for (i = 0; i < nb_nics; i++) {
1281             NICInfo *nd = &nd_table[i];
1282 
1283             if (!nd->model) {
1284                 nd->model = g_strdup("virtio");
1285             }
1286 
1287             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1288         }
1289     }
1290 
1291     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1292     qemu_fdt_add_subnode(vms->fdt, nodename);
1293     qemu_fdt_setprop_string(vms->fdt, nodename,
1294                             "compatible", "pci-host-ecam-generic");
1295     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1296     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1297     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1298     qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1299     qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1300                            nr_pcie_buses - 1);
1301     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1302 
1303     if (vms->msi_phandle) {
1304         qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1305                                vms->msi_phandle);
1306     }
1307 
1308     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1309                                  2, base_ecam, 2, size_ecam);
1310 
1311     if (vms->highmem) {
1312         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1313                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1314                                      2, base_pio, 2, size_pio,
1315                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1316                                      2, base_mmio, 2, size_mmio,
1317                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1318                                      2, base_mmio_high,
1319                                      2, base_mmio_high, 2, size_mmio_high);
1320     } else {
1321         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1322                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1323                                      2, base_pio, 2, size_pio,
1324                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1325                                      2, base_mmio, 2, size_mmio);
1326     }
1327 
1328     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1329     create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1330 
1331     if (vms->iommu) {
1332         vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1333 
1334         switch (vms->iommu) {
1335         case VIRT_IOMMU_SMMUV3:
1336             create_smmu(vms, pci->bus);
1337             qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1338                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1339             break;
1340         default:
1341             g_assert_not_reached();
1342         }
1343     }
1344 }
1345 
1346 static void create_platform_bus(VirtMachineState *vms)
1347 {
1348     DeviceState *dev;
1349     SysBusDevice *s;
1350     int i;
1351     MemoryRegion *sysmem = get_system_memory();
1352 
1353     dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1354     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1355     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1356     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1357     qdev_init_nofail(dev);
1358     vms->platform_bus_dev = dev;
1359 
1360     s = SYS_BUS_DEVICE(dev);
1361     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1362         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1363         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1364     }
1365 
1366     memory_region_add_subregion(sysmem,
1367                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1368                                 sysbus_mmio_get_region(s, 0));
1369 }
1370 
1371 static void create_secure_ram(VirtMachineState *vms,
1372                               MemoryRegion *secure_sysmem)
1373 {
1374     MemoryRegion *secram = g_new(MemoryRegion, 1);
1375     char *nodename;
1376     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1377     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1378 
1379     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1380                            &error_fatal);
1381     memory_region_add_subregion(secure_sysmem, base, secram);
1382 
1383     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1384     qemu_fdt_add_subnode(vms->fdt, nodename);
1385     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1386     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1387     qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1388     qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1389 
1390     g_free(nodename);
1391 }
1392 
1393 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1394 {
1395     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1396                                                  bootinfo);
1397 
1398     *fdt_size = board->fdt_size;
1399     return board->fdt;
1400 }
1401 
1402 static void virt_build_smbios(VirtMachineState *vms)
1403 {
1404     MachineClass *mc = MACHINE_GET_CLASS(vms);
1405     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1406     uint8_t *smbios_tables, *smbios_anchor;
1407     size_t smbios_tables_len, smbios_anchor_len;
1408     const char *product = "QEMU Virtual Machine";
1409 
1410     if (kvm_enabled()) {
1411         product = "KVM Virtual Machine";
1412     }
1413 
1414     smbios_set_defaults("QEMU", product,
1415                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1416                         true, SMBIOS_ENTRY_POINT_30);
1417 
1418     smbios_get_tables(MACHINE(vms), NULL, 0, &smbios_tables, &smbios_tables_len,
1419                       &smbios_anchor, &smbios_anchor_len);
1420 
1421     if (smbios_anchor) {
1422         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1423                         smbios_tables, smbios_tables_len);
1424         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1425                         smbios_anchor, smbios_anchor_len);
1426     }
1427 }
1428 
1429 static
1430 void virt_machine_done(Notifier *notifier, void *data)
1431 {
1432     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1433                                          machine_done);
1434     MachineState *ms = MACHINE(vms);
1435     ARMCPU *cpu = ARM_CPU(first_cpu);
1436     struct arm_boot_info *info = &vms->bootinfo;
1437     AddressSpace *as = arm_boot_address_space(cpu, info);
1438 
1439     /*
1440      * If the user provided a dtb, we assume the dynamic sysbus nodes
1441      * already are integrated there. This corresponds to a use case where
1442      * the dynamic sysbus nodes are complex and their generation is not yet
1443      * supported. In that case the user can take charge of the guest dt
1444      * while qemu takes charge of the qom stuff.
1445      */
1446     if (info->dtb_filename == NULL) {
1447         platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1448                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1449                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1450                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1451     }
1452     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1453         exit(1);
1454     }
1455 
1456     virt_acpi_setup(vms);
1457     virt_build_smbios(vms);
1458 }
1459 
1460 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1461 {
1462     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1463     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1464 
1465     if (!vmc->disallow_affinity_adjustment) {
1466         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1467          * GIC's target-list limitations. 32-bit KVM hosts currently
1468          * always create clusters of 4 CPUs, but that is expected to
1469          * change when they gain support for gicv3. When KVM is enabled
1470          * it will override the changes we make here, therefore our
1471          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1472          * and to improve SGI efficiency.
1473          */
1474         if (vms->gic_version == VIRT_GIC_VERSION_3) {
1475             clustersz = GICV3_TARGETLIST_BITS;
1476         } else {
1477             clustersz = GIC_TARGETLIST_BITS;
1478         }
1479     }
1480     return arm_cpu_mp_affinity(idx, clustersz);
1481 }
1482 
1483 static void virt_set_memmap(VirtMachineState *vms)
1484 {
1485     MachineState *ms = MACHINE(vms);
1486     hwaddr base, device_memory_base, device_memory_size;
1487     int i;
1488 
1489     vms->memmap = extended_memmap;
1490 
1491     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1492         vms->memmap[i] = base_memmap[i];
1493     }
1494 
1495     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1496         error_report("unsupported number of memory slots: %"PRIu64,
1497                      ms->ram_slots);
1498         exit(EXIT_FAILURE);
1499     }
1500 
1501     /*
1502      * We compute the base of the high IO region depending on the
1503      * amount of initial and device memory. The device memory start/size
1504      * is aligned on 1GiB. We never put the high IO region below 256GiB
1505      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1506      * The device region size assumes 1GiB page max alignment per slot.
1507      */
1508     device_memory_base =
1509         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1510     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1511 
1512     /* Base address of the high IO region */
1513     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1514     if (base < device_memory_base) {
1515         error_report("maxmem/slots too huge");
1516         exit(EXIT_FAILURE);
1517     }
1518     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1519         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1520     }
1521 
1522     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1523         hwaddr size = extended_memmap[i].size;
1524 
1525         base = ROUND_UP(base, size);
1526         vms->memmap[i].base = base;
1527         vms->memmap[i].size = size;
1528         base += size;
1529     }
1530     vms->highest_gpa = base - 1;
1531     if (device_memory_size > 0) {
1532         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1533         ms->device_memory->base = device_memory_base;
1534         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1535                            "device-memory", device_memory_size);
1536     }
1537 }
1538 
1539 /*
1540  * finalize_gic_version - Determines the final gic_version
1541  * according to the gic-version property
1542  *
1543  * Default GIC type is v2
1544  */
1545 static void finalize_gic_version(VirtMachineState *vms)
1546 {
1547     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1548 
1549     if (kvm_enabled()) {
1550         int probe_bitmap;
1551 
1552         if (!kvm_irqchip_in_kernel()) {
1553             switch (vms->gic_version) {
1554             case VIRT_GIC_VERSION_HOST:
1555                 warn_report(
1556                     "gic-version=host not relevant with kernel-irqchip=off "
1557                      "as only userspace GICv2 is supported. Using v2 ...");
1558                 return;
1559             case VIRT_GIC_VERSION_MAX:
1560             case VIRT_GIC_VERSION_NOSEL:
1561                 vms->gic_version = VIRT_GIC_VERSION_2;
1562                 return;
1563             case VIRT_GIC_VERSION_2:
1564                 return;
1565             case VIRT_GIC_VERSION_3:
1566                 error_report(
1567                     "gic-version=3 is not supported with kernel-irqchip=off");
1568                 exit(1);
1569             }
1570         }
1571 
1572         probe_bitmap = kvm_arm_vgic_probe();
1573         if (!probe_bitmap) {
1574             error_report("Unable to determine GIC version supported by host");
1575             exit(1);
1576         }
1577 
1578         switch (vms->gic_version) {
1579         case VIRT_GIC_VERSION_HOST:
1580         case VIRT_GIC_VERSION_MAX:
1581             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1582                 vms->gic_version = VIRT_GIC_VERSION_3;
1583             } else {
1584                 vms->gic_version = VIRT_GIC_VERSION_2;
1585             }
1586             return;
1587         case VIRT_GIC_VERSION_NOSEL:
1588             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1589                 vms->gic_version = VIRT_GIC_VERSION_2;
1590             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1591                 /*
1592                  * in case the host does not support v2 in-kernel emulation or
1593                  * the end-user requested more than 8 VCPUs we now default
1594                  * to v3. In any case defaulting to v2 would be broken.
1595                  */
1596                 vms->gic_version = VIRT_GIC_VERSION_3;
1597             } else if (max_cpus > GIC_NCPU) {
1598                 error_report("host only supports in-kernel GICv2 emulation "
1599                              "but more than 8 vcpus are requested");
1600                 exit(1);
1601             }
1602             break;
1603         case VIRT_GIC_VERSION_2:
1604         case VIRT_GIC_VERSION_3:
1605             break;
1606         }
1607 
1608         /* Check chosen version is effectively supported by the host */
1609         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1610             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1611             error_report("host does not support in-kernel GICv2 emulation");
1612             exit(1);
1613         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1614                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1615             error_report("host does not support in-kernel GICv3 emulation");
1616             exit(1);
1617         }
1618         return;
1619     }
1620 
1621     /* TCG mode */
1622     switch (vms->gic_version) {
1623     case VIRT_GIC_VERSION_NOSEL:
1624         vms->gic_version = VIRT_GIC_VERSION_2;
1625         break;
1626     case VIRT_GIC_VERSION_MAX:
1627         vms->gic_version = VIRT_GIC_VERSION_3;
1628         break;
1629     case VIRT_GIC_VERSION_HOST:
1630         error_report("gic-version=host requires KVM");
1631         exit(1);
1632     case VIRT_GIC_VERSION_2:
1633     case VIRT_GIC_VERSION_3:
1634         break;
1635     }
1636 }
1637 
1638 static void machvirt_init(MachineState *machine)
1639 {
1640     VirtMachineState *vms = VIRT_MACHINE(machine);
1641     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1642     MachineClass *mc = MACHINE_GET_CLASS(machine);
1643     const CPUArchIdList *possible_cpus;
1644     MemoryRegion *sysmem = get_system_memory();
1645     MemoryRegion *secure_sysmem = NULL;
1646     int n, virt_max_cpus;
1647     bool firmware_loaded;
1648     bool aarch64 = true;
1649     bool has_ged = !vmc->no_ged;
1650     unsigned int smp_cpus = machine->smp.cpus;
1651     unsigned int max_cpus = machine->smp.max_cpus;
1652 
1653     /*
1654      * In accelerated mode, the memory map is computed earlier in kvm_type()
1655      * to create a VM with the right number of IPA bits.
1656      */
1657     if (!vms->memmap) {
1658         virt_set_memmap(vms);
1659     }
1660 
1661     /* We can probe only here because during property set
1662      * KVM is not available yet
1663      */
1664     finalize_gic_version(vms);
1665 
1666     if (!cpu_type_valid(machine->cpu_type)) {
1667         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1668         exit(1);
1669     }
1670 
1671     if (vms->secure) {
1672         if (kvm_enabled()) {
1673             error_report("mach-virt: KVM does not support Security extensions");
1674             exit(1);
1675         }
1676 
1677         /*
1678          * The Secure view of the world is the same as the NonSecure,
1679          * but with a few extra devices. Create it as a container region
1680          * containing the system memory at low priority; any secure-only
1681          * devices go in at higher priority and take precedence.
1682          */
1683         secure_sysmem = g_new(MemoryRegion, 1);
1684         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1685                            UINT64_MAX);
1686         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1687     }
1688 
1689     firmware_loaded = virt_firmware_init(vms, sysmem,
1690                                          secure_sysmem ?: sysmem);
1691 
1692     /* If we have an EL3 boot ROM then the assumption is that it will
1693      * implement PSCI itself, so disable QEMU's internal implementation
1694      * so it doesn't get in the way. Instead of starting secondary
1695      * CPUs in PSCI powerdown state we will start them all running and
1696      * let the boot ROM sort them out.
1697      * The usual case is that we do use QEMU's PSCI implementation;
1698      * if the guest has EL2 then we will use SMC as the conduit,
1699      * and otherwise we will use HVC (for backwards compatibility and
1700      * because if we're using KVM then we must use HVC).
1701      */
1702     if (vms->secure && firmware_loaded) {
1703         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1704     } else if (vms->virt) {
1705         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1706     } else {
1707         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1708     }
1709 
1710     /* The maximum number of CPUs depends on the GIC version, or on how
1711      * many redistributors we can fit into the memory map.
1712      */
1713     if (vms->gic_version == VIRT_GIC_VERSION_3) {
1714         virt_max_cpus =
1715             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1716         virt_max_cpus +=
1717             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1718     } else {
1719         virt_max_cpus = GIC_NCPU;
1720     }
1721 
1722     if (max_cpus > virt_max_cpus) {
1723         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1724                      "supported by machine 'mach-virt' (%d)",
1725                      max_cpus, virt_max_cpus);
1726         exit(1);
1727     }
1728 
1729     vms->smp_cpus = smp_cpus;
1730 
1731     if (vms->virt && kvm_enabled()) {
1732         error_report("mach-virt: KVM does not support providing "
1733                      "Virtualization extensions to the guest CPU");
1734         exit(1);
1735     }
1736 
1737     create_fdt(vms);
1738 
1739     possible_cpus = mc->possible_cpu_arch_ids(machine);
1740     for (n = 0; n < possible_cpus->len; n++) {
1741         Object *cpuobj;
1742         CPUState *cs;
1743 
1744         if (n >= smp_cpus) {
1745             break;
1746         }
1747 
1748         cpuobj = object_new(possible_cpus->cpus[n].type);
1749         object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
1750                                 "mp-affinity", NULL);
1751 
1752         cs = CPU(cpuobj);
1753         cs->cpu_index = n;
1754 
1755         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1756                           &error_fatal);
1757 
1758         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1759 
1760         if (!vms->secure) {
1761             object_property_set_bool(cpuobj, false, "has_el3", NULL);
1762         }
1763 
1764         if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
1765             object_property_set_bool(cpuobj, false, "has_el2", NULL);
1766         }
1767 
1768         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1769             object_property_set_int(cpuobj, vms->psci_conduit,
1770                                     "psci-conduit", NULL);
1771 
1772             /* Secondary CPUs start in PSCI powered-down state */
1773             if (n > 0) {
1774                 object_property_set_bool(cpuobj, true,
1775                                          "start-powered-off", NULL);
1776             }
1777         }
1778 
1779         if (vmc->kvm_no_adjvtime &&
1780             object_property_find(cpuobj, "kvm-no-adjvtime", NULL)) {
1781             object_property_set_bool(cpuobj, true, "kvm-no-adjvtime", NULL);
1782         }
1783 
1784         if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1785             object_property_set_bool(cpuobj, false, "pmu", NULL);
1786         }
1787 
1788         if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1789             object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
1790                                     "reset-cbar", &error_abort);
1791         }
1792 
1793         object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1794                                  &error_abort);
1795         if (vms->secure) {
1796             object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1797                                      "secure-memory", &error_abort);
1798         }
1799 
1800         object_property_set_bool(cpuobj, true, "realized", &error_fatal);
1801         object_unref(cpuobj);
1802     }
1803     fdt_add_timer_nodes(vms);
1804     fdt_add_cpu_nodes(vms);
1805 
1806    if (!kvm_enabled()) {
1807         ARMCPU *cpu = ARM_CPU(first_cpu);
1808         bool aarch64 = object_property_get_bool(OBJECT(cpu), "aarch64", NULL);
1809 
1810         if (aarch64 && vms->highmem) {
1811             int requested_pa_size, pamax = arm_pamax(cpu);
1812 
1813             requested_pa_size = 64 - clz64(vms->highest_gpa);
1814             if (pamax < requested_pa_size) {
1815                 error_report("VCPU supports less PA bits (%d) than requested "
1816                             "by the memory map (%d)", pamax, requested_pa_size);
1817                 exit(1);
1818             }
1819         }
1820     }
1821 
1822     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
1823                                 machine->ram);
1824     if (machine->device_memory) {
1825         memory_region_add_subregion(sysmem, machine->device_memory->base,
1826                                     &machine->device_memory->mr);
1827     }
1828 
1829     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
1830 
1831     create_gic(vms);
1832 
1833     fdt_add_pmu_nodes(vms);
1834 
1835     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
1836 
1837     if (vms->secure) {
1838         create_secure_ram(vms, secure_sysmem);
1839         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
1840     }
1841 
1842     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
1843 
1844     create_rtc(vms);
1845 
1846     create_pcie(vms);
1847 
1848     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
1849         vms->acpi_dev = create_acpi_ged(vms);
1850     } else {
1851         create_gpio(vms);
1852     }
1853 
1854      /* connect powerdown request */
1855      vms->powerdown_notifier.notify = virt_powerdown_req;
1856      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
1857 
1858     /* Create mmio transports, so the user can create virtio backends
1859      * (which will be automatically plugged in to the transports). If
1860      * no backend is created the transport will just sit harmlessly idle.
1861      */
1862     create_virtio_devices(vms);
1863 
1864     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1865     rom_set_fw(vms->fw_cfg);
1866 
1867     create_platform_bus(vms);
1868 
1869     vms->bootinfo.ram_size = machine->ram_size;
1870     vms->bootinfo.nb_cpus = smp_cpus;
1871     vms->bootinfo.board_id = -1;
1872     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1873     vms->bootinfo.get_dtb = machvirt_dtb;
1874     vms->bootinfo.skip_dtb_autoload = true;
1875     vms->bootinfo.firmware_loaded = firmware_loaded;
1876     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
1877 
1878     vms->machine_done.notify = virt_machine_done;
1879     qemu_add_machine_init_done_notifier(&vms->machine_done);
1880 }
1881 
1882 static bool virt_get_secure(Object *obj, Error **errp)
1883 {
1884     VirtMachineState *vms = VIRT_MACHINE(obj);
1885 
1886     return vms->secure;
1887 }
1888 
1889 static void virt_set_secure(Object *obj, bool value, Error **errp)
1890 {
1891     VirtMachineState *vms = VIRT_MACHINE(obj);
1892 
1893     vms->secure = value;
1894 }
1895 
1896 static bool virt_get_virt(Object *obj, Error **errp)
1897 {
1898     VirtMachineState *vms = VIRT_MACHINE(obj);
1899 
1900     return vms->virt;
1901 }
1902 
1903 static void virt_set_virt(Object *obj, bool value, Error **errp)
1904 {
1905     VirtMachineState *vms = VIRT_MACHINE(obj);
1906 
1907     vms->virt = value;
1908 }
1909 
1910 static bool virt_get_highmem(Object *obj, Error **errp)
1911 {
1912     VirtMachineState *vms = VIRT_MACHINE(obj);
1913 
1914     return vms->highmem;
1915 }
1916 
1917 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1918 {
1919     VirtMachineState *vms = VIRT_MACHINE(obj);
1920 
1921     vms->highmem = value;
1922 }
1923 
1924 static bool virt_get_its(Object *obj, Error **errp)
1925 {
1926     VirtMachineState *vms = VIRT_MACHINE(obj);
1927 
1928     return vms->its;
1929 }
1930 
1931 static void virt_set_its(Object *obj, bool value, Error **errp)
1932 {
1933     VirtMachineState *vms = VIRT_MACHINE(obj);
1934 
1935     vms->its = value;
1936 }
1937 
1938 bool virt_is_acpi_enabled(VirtMachineState *vms)
1939 {
1940     if (vms->acpi == ON_OFF_AUTO_OFF) {
1941         return false;
1942     }
1943     return true;
1944 }
1945 
1946 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
1947                           void *opaque, Error **errp)
1948 {
1949     VirtMachineState *vms = VIRT_MACHINE(obj);
1950     OnOffAuto acpi = vms->acpi;
1951 
1952     visit_type_OnOffAuto(v, name, &acpi, errp);
1953 }
1954 
1955 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
1956                           void *opaque, Error **errp)
1957 {
1958     VirtMachineState *vms = VIRT_MACHINE(obj);
1959 
1960     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
1961 }
1962 
1963 static char *virt_get_gic_version(Object *obj, Error **errp)
1964 {
1965     VirtMachineState *vms = VIRT_MACHINE(obj);
1966     const char *val = vms->gic_version == VIRT_GIC_VERSION_3 ? "3" : "2";
1967 
1968     return g_strdup(val);
1969 }
1970 
1971 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1972 {
1973     VirtMachineState *vms = VIRT_MACHINE(obj);
1974 
1975     if (!strcmp(value, "3")) {
1976         vms->gic_version = VIRT_GIC_VERSION_3;
1977     } else if (!strcmp(value, "2")) {
1978         vms->gic_version = VIRT_GIC_VERSION_2;
1979     } else if (!strcmp(value, "host")) {
1980         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
1981     } else if (!strcmp(value, "max")) {
1982         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
1983     } else {
1984         error_setg(errp, "Invalid gic-version value");
1985         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
1986     }
1987 }
1988 
1989 static char *virt_get_iommu(Object *obj, Error **errp)
1990 {
1991     VirtMachineState *vms = VIRT_MACHINE(obj);
1992 
1993     switch (vms->iommu) {
1994     case VIRT_IOMMU_NONE:
1995         return g_strdup("none");
1996     case VIRT_IOMMU_SMMUV3:
1997         return g_strdup("smmuv3");
1998     default:
1999         g_assert_not_reached();
2000     }
2001 }
2002 
2003 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2004 {
2005     VirtMachineState *vms = VIRT_MACHINE(obj);
2006 
2007     if (!strcmp(value, "smmuv3")) {
2008         vms->iommu = VIRT_IOMMU_SMMUV3;
2009     } else if (!strcmp(value, "none")) {
2010         vms->iommu = VIRT_IOMMU_NONE;
2011     } else {
2012         error_setg(errp, "Invalid iommu value");
2013         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2014     }
2015 }
2016 
2017 static CpuInstanceProperties
2018 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2019 {
2020     MachineClass *mc = MACHINE_GET_CLASS(ms);
2021     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2022 
2023     assert(cpu_index < possible_cpus->len);
2024     return possible_cpus->cpus[cpu_index].props;
2025 }
2026 
2027 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2028 {
2029     return idx % ms->numa_state->num_nodes;
2030 }
2031 
2032 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2033 {
2034     int n;
2035     unsigned int max_cpus = ms->smp.max_cpus;
2036     VirtMachineState *vms = VIRT_MACHINE(ms);
2037 
2038     if (ms->possible_cpus) {
2039         assert(ms->possible_cpus->len == max_cpus);
2040         return ms->possible_cpus;
2041     }
2042 
2043     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2044                                   sizeof(CPUArchId) * max_cpus);
2045     ms->possible_cpus->len = max_cpus;
2046     for (n = 0; n < ms->possible_cpus->len; n++) {
2047         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2048         ms->possible_cpus->cpus[n].arch_id =
2049             virt_cpu_mp_affinity(vms, n);
2050         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2051         ms->possible_cpus->cpus[n].props.thread_id = n;
2052     }
2053     return ms->possible_cpus;
2054 }
2055 
2056 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2057                                  Error **errp)
2058 {
2059     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2060     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2061 
2062     if (is_nvdimm) {
2063         error_setg(errp, "nvdimm is not yet supported");
2064         return;
2065     }
2066 
2067     if (!vms->acpi_dev) {
2068         error_setg(errp,
2069                    "memory hotplug is not enabled: missing acpi-ged device");
2070         return;
2071     }
2072 
2073     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2074 }
2075 
2076 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2077                              DeviceState *dev, Error **errp)
2078 {
2079     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2080     Error *local_err = NULL;
2081 
2082     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms), &local_err);
2083     if (local_err) {
2084         goto out;
2085     }
2086 
2087     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2088                          dev, &error_abort);
2089 
2090 out:
2091     error_propagate(errp, local_err);
2092 }
2093 
2094 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2095                                             DeviceState *dev, Error **errp)
2096 {
2097     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2098         virt_memory_pre_plug(hotplug_dev, dev, errp);
2099     }
2100 }
2101 
2102 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2103                                         DeviceState *dev, Error **errp)
2104 {
2105     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2106 
2107     if (vms->platform_bus_dev) {
2108         if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
2109             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2110                                      SYS_BUS_DEVICE(dev));
2111         }
2112     }
2113     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2114         virt_memory_plug(hotplug_dev, dev, errp);
2115     }
2116     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2117         PCIDevice *pdev = PCI_DEVICE(dev);
2118 
2119         vms->iommu = VIRT_IOMMU_VIRTIO;
2120         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2121         create_virtio_iommu_dt_bindings(vms, errp);
2122     }
2123 }
2124 
2125 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2126                                           DeviceState *dev, Error **errp)
2127 {
2128     error_setg(errp, "device unplug request for unsupported device"
2129                " type: %s", object_get_typename(OBJECT(dev)));
2130 }
2131 
2132 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2133                                                         DeviceState *dev)
2134 {
2135     if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE) ||
2136        (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
2137         return HOTPLUG_HANDLER(machine);
2138     }
2139     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2140         VirtMachineState *vms = VIRT_MACHINE(machine);
2141 
2142         if (!vms->bootinfo.firmware_loaded || !virt_is_acpi_enabled(vms)) {
2143             return HOTPLUG_HANDLER(machine);
2144         }
2145     }
2146     return NULL;
2147 }
2148 
2149 /*
2150  * for arm64 kvm_type [7-0] encodes the requested number of bits
2151  * in the IPA address space
2152  */
2153 static int virt_kvm_type(MachineState *ms, const char *type_str)
2154 {
2155     VirtMachineState *vms = VIRT_MACHINE(ms);
2156     int max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms);
2157     int requested_pa_size;
2158 
2159     /* we freeze the memory map to compute the highest gpa */
2160     virt_set_memmap(vms);
2161 
2162     requested_pa_size = 64 - clz64(vms->highest_gpa);
2163 
2164     if (requested_pa_size > max_vm_pa_size) {
2165         error_report("-m and ,maxmem option values "
2166                      "require an IPA range (%d bits) larger than "
2167                      "the one supported by the host (%d bits)",
2168                      requested_pa_size, max_vm_pa_size);
2169        exit(1);
2170     }
2171     /*
2172      * By default we return 0 which corresponds to an implicit legacy
2173      * 40b IPA setting. Otherwise we return the actual requested PA
2174      * logsize
2175      */
2176     return requested_pa_size > 40 ? requested_pa_size : 0;
2177 }
2178 
2179 static void virt_machine_class_init(ObjectClass *oc, void *data)
2180 {
2181     MachineClass *mc = MACHINE_CLASS(oc);
2182     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2183 
2184     mc->init = machvirt_init;
2185     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2186      * The value may be reduced later when we have more information about the
2187      * configuration of the particular instance.
2188      */
2189     mc->max_cpus = 512;
2190     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2191     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2192     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2193     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2194     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2195     mc->block_default_type = IF_VIRTIO;
2196     mc->no_cdrom = 1;
2197     mc->pci_allow_0_address = true;
2198     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2199     mc->minimum_page_bits = 12;
2200     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2201     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2202     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2203     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2204     mc->kvm_type = virt_kvm_type;
2205     assert(!mc->get_hotplug_handler);
2206     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2207     hc->pre_plug = virt_machine_device_pre_plug_cb;
2208     hc->plug = virt_machine_device_plug_cb;
2209     hc->unplug_request = virt_machine_device_unplug_request_cb;
2210     mc->numa_mem_supported = true;
2211     mc->auto_enable_numa_with_memhp = true;
2212     mc->default_ram_id = "mach-virt.ram";
2213 
2214     object_class_property_add(oc, "acpi", "OnOffAuto",
2215         virt_get_acpi, virt_set_acpi,
2216         NULL, NULL, &error_abort);
2217     object_class_property_set_description(oc, "acpi",
2218         "Enable ACPI", &error_abort);
2219 }
2220 
2221 static void virt_instance_init(Object *obj)
2222 {
2223     VirtMachineState *vms = VIRT_MACHINE(obj);
2224     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2225 
2226     /* EL3 is disabled by default on virt: this makes us consistent
2227      * between KVM and TCG for this board, and it also allows us to
2228      * boot UEFI blobs which assume no TrustZone support.
2229      */
2230     vms->secure = false;
2231     object_property_add_bool(obj, "secure", virt_get_secure,
2232                              virt_set_secure, NULL);
2233     object_property_set_description(obj, "secure",
2234                                     "Set on/off to enable/disable the ARM "
2235                                     "Security Extensions (TrustZone)",
2236                                     NULL);
2237 
2238     /* EL2 is also disabled by default, for similar reasons */
2239     vms->virt = false;
2240     object_property_add_bool(obj, "virtualization", virt_get_virt,
2241                              virt_set_virt, NULL);
2242     object_property_set_description(obj, "virtualization",
2243                                     "Set on/off to enable/disable emulating a "
2244                                     "guest CPU which implements the ARM "
2245                                     "Virtualization Extensions",
2246                                     NULL);
2247 
2248     /* High memory is enabled by default */
2249     vms->highmem = true;
2250     object_property_add_bool(obj, "highmem", virt_get_highmem,
2251                              virt_set_highmem, NULL);
2252     object_property_set_description(obj, "highmem",
2253                                     "Set on/off to enable/disable using "
2254                                     "physical address space above 32 bits",
2255                                     NULL);
2256     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
2257     object_property_add_str(obj, "gic-version", virt_get_gic_version,
2258                         virt_set_gic_version, NULL);
2259     object_property_set_description(obj, "gic-version",
2260                                     "Set GIC version. "
2261                                     "Valid values are 2, 3, host and max",
2262                                     NULL);
2263 
2264     vms->highmem_ecam = !vmc->no_highmem_ecam;
2265 
2266     if (vmc->no_its) {
2267         vms->its = false;
2268     } else {
2269         /* Default allows ITS instantiation */
2270         vms->its = true;
2271         object_property_add_bool(obj, "its", virt_get_its,
2272                                  virt_set_its, NULL);
2273         object_property_set_description(obj, "its",
2274                                         "Set on/off to enable/disable "
2275                                         "ITS instantiation",
2276                                         NULL);
2277     }
2278 
2279     /* Default disallows iommu instantiation */
2280     vms->iommu = VIRT_IOMMU_NONE;
2281     object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu, NULL);
2282     object_property_set_description(obj, "iommu",
2283                                     "Set the IOMMU type. "
2284                                     "Valid values are none and smmuv3",
2285                                     NULL);
2286 
2287     vms->irqmap = a15irqmap;
2288 
2289     virt_flash_create(vms);
2290 }
2291 
2292 static const TypeInfo virt_machine_info = {
2293     .name          = TYPE_VIRT_MACHINE,
2294     .parent        = TYPE_MACHINE,
2295     .abstract      = true,
2296     .instance_size = sizeof(VirtMachineState),
2297     .class_size    = sizeof(VirtMachineClass),
2298     .class_init    = virt_machine_class_init,
2299     .instance_init = virt_instance_init,
2300     .interfaces = (InterfaceInfo[]) {
2301          { TYPE_HOTPLUG_HANDLER },
2302          { }
2303     },
2304 };
2305 
2306 static void machvirt_machine_init(void)
2307 {
2308     type_register_static(&virt_machine_info);
2309 }
2310 type_init(machvirt_machine_init);
2311 
2312 static void virt_machine_5_0_options(MachineClass *mc)
2313 {
2314     static GlobalProperty compat[] = {
2315         { TYPE_TPM_TIS_SYSBUS, "ppi", "false" },
2316     };
2317 
2318     compat_props_add(mc->compat_props, compat, G_N_ELEMENTS(compat));
2319 }
2320 DEFINE_VIRT_MACHINE_AS_LATEST(5, 0)
2321 
2322 static void virt_machine_4_2_options(MachineClass *mc)
2323 {
2324     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2325 
2326     virt_machine_5_0_options(mc);
2327     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
2328     vmc->kvm_no_adjvtime = true;
2329 }
2330 DEFINE_VIRT_MACHINE(4, 2)
2331 
2332 static void virt_machine_4_1_options(MachineClass *mc)
2333 {
2334     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2335 
2336     virt_machine_4_2_options(mc);
2337     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2338     vmc->no_ged = true;
2339     mc->auto_enable_numa_with_memhp = false;
2340 }
2341 DEFINE_VIRT_MACHINE(4, 1)
2342 
2343 static void virt_machine_4_0_options(MachineClass *mc)
2344 {
2345     virt_machine_4_1_options(mc);
2346     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2347 }
2348 DEFINE_VIRT_MACHINE(4, 0)
2349 
2350 static void virt_machine_3_1_options(MachineClass *mc)
2351 {
2352     virt_machine_4_0_options(mc);
2353     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2354 }
2355 DEFINE_VIRT_MACHINE(3, 1)
2356 
2357 static void virt_machine_3_0_options(MachineClass *mc)
2358 {
2359     virt_machine_3_1_options(mc);
2360     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2361 }
2362 DEFINE_VIRT_MACHINE(3, 0)
2363 
2364 static void virt_machine_2_12_options(MachineClass *mc)
2365 {
2366     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2367 
2368     virt_machine_3_0_options(mc);
2369     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2370     vmc->no_highmem_ecam = true;
2371     mc->max_cpus = 255;
2372 }
2373 DEFINE_VIRT_MACHINE(2, 12)
2374 
2375 static void virt_machine_2_11_options(MachineClass *mc)
2376 {
2377     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2378 
2379     virt_machine_2_12_options(mc);
2380     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2381     vmc->smbios_old_sys_ver = true;
2382 }
2383 DEFINE_VIRT_MACHINE(2, 11)
2384 
2385 static void virt_machine_2_10_options(MachineClass *mc)
2386 {
2387     virt_machine_2_11_options(mc);
2388     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2389     /* before 2.11 we never faulted accesses to bad addresses */
2390     mc->ignore_memory_transaction_failures = true;
2391 }
2392 DEFINE_VIRT_MACHINE(2, 10)
2393 
2394 static void virt_machine_2_9_options(MachineClass *mc)
2395 {
2396     virt_machine_2_10_options(mc);
2397     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2398 }
2399 DEFINE_VIRT_MACHINE(2, 9)
2400 
2401 static void virt_machine_2_8_options(MachineClass *mc)
2402 {
2403     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2404 
2405     virt_machine_2_9_options(mc);
2406     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
2407     /* For 2.8 and earlier we falsely claimed in the DT that
2408      * our timers were edge-triggered, not level-triggered.
2409      */
2410     vmc->claim_edge_triggered_timers = true;
2411 }
2412 DEFINE_VIRT_MACHINE(2, 8)
2413 
2414 static void virt_machine_2_7_options(MachineClass *mc)
2415 {
2416     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2417 
2418     virt_machine_2_8_options(mc);
2419     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2420     /* ITS was introduced with 2.8 */
2421     vmc->no_its = true;
2422     /* Stick with 1K pages for migration compatibility */
2423     mc->minimum_page_bits = 0;
2424 }
2425 DEFINE_VIRT_MACHINE(2, 7)
2426 
2427 static void virt_machine_2_6_options(MachineClass *mc)
2428 {
2429     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2430 
2431     virt_machine_2_7_options(mc);
2432     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
2433     vmc->disallow_affinity_adjustment = true;
2434     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2435     vmc->no_pmu = true;
2436 }
2437 DEFINE_VIRT_MACHINE(2, 6)
2438