xref: /openbmc/qemu/hw/arm/virt.c (revision a3fc8396352e945f9d14cac0237ebf9d91745969)
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 "qapi/error.h"
33 #include "hw/sysbus.h"
34 #include "hw/arm/arm.h"
35 #include "hw/arm/primecell.h"
36 #include "hw/arm/virt.h"
37 #include "hw/vfio/vfio-calxeda-xgmac.h"
38 #include "hw/vfio/vfio-amd-xgbe.h"
39 #include "hw/devices.h"
40 #include "net/net.h"
41 #include "sysemu/block-backend.h"
42 #include "sysemu/device_tree.h"
43 #include "sysemu/numa.h"
44 #include "sysemu/sysemu.h"
45 #include "sysemu/kvm.h"
46 #include "hw/compat.h"
47 #include "hw/loader.h"
48 #include "exec/address-spaces.h"
49 #include "qemu/bitops.h"
50 #include "qemu/error-report.h"
51 #include "hw/pci-host/gpex.h"
52 #include "hw/arm/sysbus-fdt.h"
53 #include "hw/platform-bus.h"
54 #include "hw/arm/fdt.h"
55 #include "hw/intc/arm_gic.h"
56 #include "hw/intc/arm_gicv3_common.h"
57 #include "kvm_arm.h"
58 #include "hw/smbios/smbios.h"
59 #include "qapi/visitor.h"
60 #include "standard-headers/linux/input.h"
61 #include "hw/arm/smmuv3.h"
62 
63 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
64     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
65                                                     void *data) \
66     { \
67         MachineClass *mc = MACHINE_CLASS(oc); \
68         virt_machine_##major##_##minor##_options(mc); \
69         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
70         if (latest) { \
71             mc->alias = "virt"; \
72         } \
73     } \
74     static const TypeInfo machvirt_##major##_##minor##_info = { \
75         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
76         .parent = TYPE_VIRT_MACHINE, \
77         .instance_init = virt_##major##_##minor##_instance_init, \
78         .class_init = virt_##major##_##minor##_class_init, \
79     }; \
80     static void machvirt_machine_##major##_##minor##_init(void) \
81     { \
82         type_register_static(&machvirt_##major##_##minor##_info); \
83     } \
84     type_init(machvirt_machine_##major##_##minor##_init);
85 
86 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
87     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
88 #define DEFINE_VIRT_MACHINE(major, minor) \
89     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
90 
91 
92 /* Number of external interrupt lines to configure the GIC with */
93 #define NUM_IRQS 256
94 
95 #define PLATFORM_BUS_NUM_IRQS 64
96 
97 static ARMPlatformBusSystemParams platform_bus_params;
98 
99 /* RAM limit in GB. Since VIRT_MEM starts at the 1GB mark, this means
100  * RAM can go up to the 256GB mark, leaving 256GB of the physical
101  * address space unallocated and free for future use between 256G and 512G.
102  * If we need to provide more RAM to VMs in the future then we need to:
103  *  * allocate a second bank of RAM starting at 2TB and working up
104  *  * fix the DT and ACPI table generation code in QEMU to correctly
105  *    report two split lumps of RAM to the guest
106  *  * fix KVM in the host kernel to allow guests with >40 bit address spaces
107  * (We don't want to fill all the way up to 512GB with RAM because
108  * we might want it for non-RAM purposes later. Conversely it seems
109  * reasonable to assume that anybody configuring a VM with a quarter
110  * of a terabyte of RAM will be doing it on a host with more than a
111  * terabyte of physical address space.)
112  */
113 #define RAMLIMIT_GB 255
114 #define RAMLIMIT_BYTES (RAMLIMIT_GB * 1024ULL * 1024 * 1024)
115 
116 /* Addresses and sizes of our components.
117  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
118  * 128MB..256MB is used for miscellaneous device I/O.
119  * 256MB..1GB is reserved for possible future PCI support (ie where the
120  * PCI memory window will go if we add a PCI host controller).
121  * 1GB and up is RAM (which may happily spill over into the
122  * high memory region beyond 4GB).
123  * This represents a compromise between how much RAM can be given to
124  * a 32 bit VM and leaving space for expansion and in particular for PCI.
125  * Note that devices should generally be placed at multiples of 0x10000,
126  * to accommodate guests using 64K pages.
127  */
128 static const MemMapEntry a15memmap[] = {
129     /* Space up to 0x8000000 is reserved for a boot ROM */
130     [VIRT_FLASH] =              {          0, 0x08000000 },
131     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
132     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
133     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
134     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
135     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
136     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
137     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
138     /* This redistributor space allows up to 2*64kB*123 CPUs */
139     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
140     [VIRT_UART] =               { 0x09000000, 0x00001000 },
141     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
142     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
143     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
144     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
145     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
146     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
147     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
148     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
149     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
150     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
151     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
152     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
153     [VIRT_MEM] =                { 0x40000000, RAMLIMIT_BYTES },
154     /* Second PCIe window, 512GB wide at the 512GB boundary */
155     [VIRT_PCIE_MMIO_HIGH] =   { 0x8000000000ULL, 0x8000000000ULL },
156 };
157 
158 static const int a15irqmap[] = {
159     [VIRT_UART] = 1,
160     [VIRT_RTC] = 2,
161     [VIRT_PCIE] = 3, /* ... to 6 */
162     [VIRT_GPIO] = 7,
163     [VIRT_SECURE_UART] = 8,
164     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
165     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
166     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
167     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
168 };
169 
170 static const char *valid_cpus[] = {
171     ARM_CPU_TYPE_NAME("cortex-a15"),
172     ARM_CPU_TYPE_NAME("cortex-a53"),
173     ARM_CPU_TYPE_NAME("cortex-a57"),
174     ARM_CPU_TYPE_NAME("host"),
175     ARM_CPU_TYPE_NAME("max"),
176 };
177 
178 static bool cpu_type_valid(const char *cpu)
179 {
180     int i;
181 
182     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
183         if (strcmp(cpu, valid_cpus[i]) == 0) {
184             return true;
185         }
186     }
187     return false;
188 }
189 
190 static void create_fdt(VirtMachineState *vms)
191 {
192     void *fdt = create_device_tree(&vms->fdt_size);
193 
194     if (!fdt) {
195         error_report("create_device_tree() failed");
196         exit(1);
197     }
198 
199     vms->fdt = fdt;
200 
201     /* Header */
202     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
203     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
204     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
205 
206     /*
207      * /chosen and /memory nodes must exist for load_dtb
208      * to fill in necessary properties later
209      */
210     qemu_fdt_add_subnode(fdt, "/chosen");
211     qemu_fdt_add_subnode(fdt, "/memory");
212     qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory");
213 
214     /* Clock node, for the benefit of the UART. The kernel device tree
215      * binding documentation claims the PL011 node clock properties are
216      * optional but in practice if you omit them the kernel refuses to
217      * probe for the device.
218      */
219     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
220     qemu_fdt_add_subnode(fdt, "/apb-pclk");
221     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
222     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
223     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
224     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
225                                 "clk24mhz");
226     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
227 
228     if (have_numa_distance) {
229         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
230         uint32_t *matrix = g_malloc0(size);
231         int idx, i, j;
232 
233         for (i = 0; i < nb_numa_nodes; i++) {
234             for (j = 0; j < nb_numa_nodes; j++) {
235                 idx = (i * nb_numa_nodes + j) * 3;
236                 matrix[idx + 0] = cpu_to_be32(i);
237                 matrix[idx + 1] = cpu_to_be32(j);
238                 matrix[idx + 2] = cpu_to_be32(numa_info[i].distance[j]);
239             }
240         }
241 
242         qemu_fdt_add_subnode(fdt, "/distance-map");
243         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
244                                 "numa-distance-map-v1");
245         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
246                          matrix, size);
247         g_free(matrix);
248     }
249 }
250 
251 static void fdt_add_timer_nodes(const VirtMachineState *vms)
252 {
253     /* On real hardware these interrupts are level-triggered.
254      * On KVM they were edge-triggered before host kernel version 4.4,
255      * and level-triggered afterwards.
256      * On emulated QEMU they are level-triggered.
257      *
258      * Getting the DTB info about them wrong is awkward for some
259      * guest kernels:
260      *  pre-4.8 ignore the DT and leave the interrupt configured
261      *   with whatever the GIC reset value (or the bootloader) left it at
262      *  4.8 before rc6 honour the incorrect data by programming it back
263      *   into the GIC, causing problems
264      *  4.8rc6 and later ignore the DT and always write "level triggered"
265      *   into the GIC
266      *
267      * For backwards-compatibility, virt-2.8 and earlier will continue
268      * to say these are edge-triggered, but later machines will report
269      * the correct information.
270      */
271     ARMCPU *armcpu;
272     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
273     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
274 
275     if (vmc->claim_edge_triggered_timers) {
276         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
277     }
278 
279     if (vms->gic_version == 2) {
280         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
281                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
282                              (1 << vms->smp_cpus) - 1);
283     }
284 
285     qemu_fdt_add_subnode(vms->fdt, "/timer");
286 
287     armcpu = ARM_CPU(qemu_get_cpu(0));
288     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
289         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
290         qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
291                          compat, sizeof(compat));
292     } else {
293         qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
294                                 "arm,armv7-timer");
295     }
296     qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
297     qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
298                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
299                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
300                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
301                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
302 }
303 
304 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
305 {
306     int cpu;
307     int addr_cells = 1;
308     const MachineState *ms = MACHINE(vms);
309 
310     /*
311      * From Documentation/devicetree/bindings/arm/cpus.txt
312      *  On ARM v8 64-bit systems value should be set to 2,
313      *  that corresponds to the MPIDR_EL1 register size.
314      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
315      *  in the system, #address-cells can be set to 1, since
316      *  MPIDR_EL1[63:32] bits are not used for CPUs
317      *  identification.
318      *
319      *  Here we actually don't know whether our system is 32- or 64-bit one.
320      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
321      *  at least one of them has Aff3 populated, we set #address-cells to 2.
322      */
323     for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
324         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
325 
326         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
327             addr_cells = 2;
328             break;
329         }
330     }
331 
332     qemu_fdt_add_subnode(vms->fdt, "/cpus");
333     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
334     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
335 
336     for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
337         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
338         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
339         CPUState *cs = CPU(armcpu);
340 
341         qemu_fdt_add_subnode(vms->fdt, nodename);
342         qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
343         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
344                                     armcpu->dtb_compatible);
345 
346         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
347             && vms->smp_cpus > 1) {
348             qemu_fdt_setprop_string(vms->fdt, nodename,
349                                         "enable-method", "psci");
350         }
351 
352         if (addr_cells == 2) {
353             qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
354                                  armcpu->mp_affinity);
355         } else {
356             qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
357                                   armcpu->mp_affinity);
358         }
359 
360         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
361             qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
362                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
363         }
364 
365         g_free(nodename);
366     }
367 }
368 
369 static void fdt_add_its_gic_node(VirtMachineState *vms)
370 {
371     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
372     qemu_fdt_add_subnode(vms->fdt, "/intc/its");
373     qemu_fdt_setprop_string(vms->fdt, "/intc/its", "compatible",
374                             "arm,gic-v3-its");
375     qemu_fdt_setprop(vms->fdt, "/intc/its", "msi-controller", NULL, 0);
376     qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/its", "reg",
377                                  2, vms->memmap[VIRT_GIC_ITS].base,
378                                  2, vms->memmap[VIRT_GIC_ITS].size);
379     qemu_fdt_setprop_cell(vms->fdt, "/intc/its", "phandle", vms->msi_phandle);
380 }
381 
382 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
383 {
384     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
385     qemu_fdt_add_subnode(vms->fdt, "/intc/v2m");
386     qemu_fdt_setprop_string(vms->fdt, "/intc/v2m", "compatible",
387                             "arm,gic-v2m-frame");
388     qemu_fdt_setprop(vms->fdt, "/intc/v2m", "msi-controller", NULL, 0);
389     qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/v2m", "reg",
390                                  2, vms->memmap[VIRT_GIC_V2M].base,
391                                  2, vms->memmap[VIRT_GIC_V2M].size);
392     qemu_fdt_setprop_cell(vms->fdt, "/intc/v2m", "phandle", vms->msi_phandle);
393 }
394 
395 static void fdt_add_gic_node(VirtMachineState *vms)
396 {
397     vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
398     qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
399 
400     qemu_fdt_add_subnode(vms->fdt, "/intc");
401     qemu_fdt_setprop_cell(vms->fdt, "/intc", "#interrupt-cells", 3);
402     qemu_fdt_setprop(vms->fdt, "/intc", "interrupt-controller", NULL, 0);
403     qemu_fdt_setprop_cell(vms->fdt, "/intc", "#address-cells", 0x2);
404     qemu_fdt_setprop_cell(vms->fdt, "/intc", "#size-cells", 0x2);
405     qemu_fdt_setprop(vms->fdt, "/intc", "ranges", NULL, 0);
406     if (vms->gic_version == 3) {
407         qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
408                                 "arm,gic-v3");
409         qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
410                                      2, vms->memmap[VIRT_GIC_DIST].base,
411                                      2, vms->memmap[VIRT_GIC_DIST].size,
412                                      2, vms->memmap[VIRT_GIC_REDIST].base,
413                                      2, vms->memmap[VIRT_GIC_REDIST].size);
414         if (vms->virt) {
415             qemu_fdt_setprop_cells(vms->fdt, "/intc", "interrupts",
416                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GICV3_MAINT_IRQ,
417                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
418         }
419     } else {
420         /* 'cortex-a15-gic' means 'GIC v2' */
421         qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
422                                 "arm,cortex-a15-gic");
423         qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
424                                       2, vms->memmap[VIRT_GIC_DIST].base,
425                                       2, vms->memmap[VIRT_GIC_DIST].size,
426                                       2, vms->memmap[VIRT_GIC_CPU].base,
427                                       2, vms->memmap[VIRT_GIC_CPU].size);
428     }
429 
430     qemu_fdt_setprop_cell(vms->fdt, "/intc", "phandle", vms->gic_phandle);
431 }
432 
433 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
434 {
435     CPUState *cpu;
436     ARMCPU *armcpu;
437     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
438 
439     CPU_FOREACH(cpu) {
440         armcpu = ARM_CPU(cpu);
441         if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
442             return;
443         }
444         if (kvm_enabled()) {
445             if (kvm_irqchip_in_kernel()) {
446                 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
447             }
448             kvm_arm_pmu_init(cpu);
449         }
450     }
451 
452     if (vms->gic_version == 2) {
453         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
454                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
455                              (1 << vms->smp_cpus) - 1);
456     }
457 
458     armcpu = ARM_CPU(qemu_get_cpu(0));
459     qemu_fdt_add_subnode(vms->fdt, "/pmu");
460     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
461         const char compat[] = "arm,armv8-pmuv3";
462         qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
463                          compat, sizeof(compat));
464         qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
465                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
466     }
467 }
468 
469 static void create_its(VirtMachineState *vms, DeviceState *gicdev)
470 {
471     const char *itsclass = its_class_name();
472     DeviceState *dev;
473 
474     if (!itsclass) {
475         /* Do nothing if not supported */
476         return;
477     }
478 
479     dev = qdev_create(NULL, itsclass);
480 
481     object_property_set_link(OBJECT(dev), OBJECT(gicdev), "parent-gicv3",
482                              &error_abort);
483     qdev_init_nofail(dev);
484     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
485 
486     fdt_add_its_gic_node(vms);
487 }
488 
489 static void create_v2m(VirtMachineState *vms, qemu_irq *pic)
490 {
491     int i;
492     int irq = vms->irqmap[VIRT_GIC_V2M];
493     DeviceState *dev;
494 
495     dev = qdev_create(NULL, "arm-gicv2m");
496     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
497     qdev_prop_set_uint32(dev, "base-spi", irq);
498     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
499     qdev_init_nofail(dev);
500 
501     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
502         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
503     }
504 
505     fdt_add_v2m_gic_node(vms);
506 }
507 
508 static void create_gic(VirtMachineState *vms, qemu_irq *pic)
509 {
510     /* We create a standalone GIC */
511     DeviceState *gicdev;
512     SysBusDevice *gicbusdev;
513     const char *gictype;
514     int type = vms->gic_version, i;
515 
516     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
517 
518     gicdev = qdev_create(NULL, gictype);
519     qdev_prop_set_uint32(gicdev, "revision", type);
520     qdev_prop_set_uint32(gicdev, "num-cpu", smp_cpus);
521     /* Note that the num-irq property counts both internal and external
522      * interrupts; there are always 32 of the former (mandated by GIC spec).
523      */
524     qdev_prop_set_uint32(gicdev, "num-irq", NUM_IRQS + 32);
525     if (!kvm_irqchip_in_kernel()) {
526         qdev_prop_set_bit(gicdev, "has-security-extensions", vms->secure);
527     }
528     qdev_init_nofail(gicdev);
529     gicbusdev = SYS_BUS_DEVICE(gicdev);
530     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
531     if (type == 3) {
532         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
533     } else {
534         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
535     }
536 
537     /* Wire the outputs from each CPU's generic timer and the GICv3
538      * maintenance interrupt signal to the appropriate GIC PPI inputs,
539      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
540      */
541     for (i = 0; i < smp_cpus; i++) {
542         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
543         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
544         int irq;
545         /* Mapping from the output timer irq lines from the CPU to the
546          * GIC PPI inputs we use for the virt board.
547          */
548         const int timer_irq[] = {
549             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
550             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
551             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
552             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
553         };
554 
555         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
556             qdev_connect_gpio_out(cpudev, irq,
557                                   qdev_get_gpio_in(gicdev,
558                                                    ppibase + timer_irq[irq]));
559         }
560 
561         qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt", 0,
562                                     qdev_get_gpio_in(gicdev, ppibase
563                                                      + ARCH_GICV3_MAINT_IRQ));
564         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
565                                     qdev_get_gpio_in(gicdev, ppibase
566                                                      + VIRTUAL_PMU_IRQ));
567 
568         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
569         sysbus_connect_irq(gicbusdev, i + smp_cpus,
570                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
571         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
572                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
573         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
574                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
575     }
576 
577     for (i = 0; i < NUM_IRQS; i++) {
578         pic[i] = qdev_get_gpio_in(gicdev, i);
579     }
580 
581     fdt_add_gic_node(vms);
582 
583     if (type == 3 && vms->its) {
584         create_its(vms, gicdev);
585     } else if (type == 2) {
586         create_v2m(vms, pic);
587     }
588 }
589 
590 static void create_uart(const VirtMachineState *vms, qemu_irq *pic, int uart,
591                         MemoryRegion *mem, Chardev *chr)
592 {
593     char *nodename;
594     hwaddr base = vms->memmap[uart].base;
595     hwaddr size = vms->memmap[uart].size;
596     int irq = vms->irqmap[uart];
597     const char compat[] = "arm,pl011\0arm,primecell";
598     const char clocknames[] = "uartclk\0apb_pclk";
599     DeviceState *dev = qdev_create(NULL, "pl011");
600     SysBusDevice *s = SYS_BUS_DEVICE(dev);
601 
602     qdev_prop_set_chr(dev, "chardev", chr);
603     qdev_init_nofail(dev);
604     memory_region_add_subregion(mem, base,
605                                 sysbus_mmio_get_region(s, 0));
606     sysbus_connect_irq(s, 0, pic[irq]);
607 
608     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
609     qemu_fdt_add_subnode(vms->fdt, nodename);
610     /* Note that we can't use setprop_string because of the embedded NUL */
611     qemu_fdt_setprop(vms->fdt, nodename, "compatible",
612                          compat, sizeof(compat));
613     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
614                                      2, base, 2, size);
615     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
616                                GIC_FDT_IRQ_TYPE_SPI, irq,
617                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
618     qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
619                                vms->clock_phandle, vms->clock_phandle);
620     qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
621                          clocknames, sizeof(clocknames));
622 
623     if (uart == VIRT_UART) {
624         qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
625     } else {
626         /* Mark as not usable by the normal world */
627         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
628         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
629     }
630 
631     g_free(nodename);
632 }
633 
634 static void create_rtc(const VirtMachineState *vms, qemu_irq *pic)
635 {
636     char *nodename;
637     hwaddr base = vms->memmap[VIRT_RTC].base;
638     hwaddr size = vms->memmap[VIRT_RTC].size;
639     int irq = vms->irqmap[VIRT_RTC];
640     const char compat[] = "arm,pl031\0arm,primecell";
641 
642     sysbus_create_simple("pl031", base, pic[irq]);
643 
644     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
645     qemu_fdt_add_subnode(vms->fdt, nodename);
646     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
647     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
648                                  2, base, 2, size);
649     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
650                            GIC_FDT_IRQ_TYPE_SPI, irq,
651                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
652     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
653     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
654     g_free(nodename);
655 }
656 
657 static DeviceState *gpio_key_dev;
658 static void virt_powerdown_req(Notifier *n, void *opaque)
659 {
660     /* use gpio Pin 3 for power button event */
661     qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
662 }
663 
664 static Notifier virt_system_powerdown_notifier = {
665     .notify = virt_powerdown_req
666 };
667 
668 static void create_gpio(const VirtMachineState *vms, qemu_irq *pic)
669 {
670     char *nodename;
671     DeviceState *pl061_dev;
672     hwaddr base = vms->memmap[VIRT_GPIO].base;
673     hwaddr size = vms->memmap[VIRT_GPIO].size;
674     int irq = vms->irqmap[VIRT_GPIO];
675     const char compat[] = "arm,pl061\0arm,primecell";
676 
677     pl061_dev = sysbus_create_simple("pl061", base, pic[irq]);
678 
679     uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
680     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
681     qemu_fdt_add_subnode(vms->fdt, nodename);
682     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
683                                  2, base, 2, size);
684     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
685     qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
686     qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
687     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
688                            GIC_FDT_IRQ_TYPE_SPI, irq,
689                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
690     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
691     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
692     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
693 
694     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
695                                         qdev_get_gpio_in(pl061_dev, 3));
696     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
697     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
698     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
699     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
700 
701     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
702     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
703                             "label", "GPIO Key Poweroff");
704     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
705                           KEY_POWER);
706     qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
707                            "gpios", phandle, 3, 0);
708 
709     /* connect powerdown request */
710     qemu_register_powerdown_notifier(&virt_system_powerdown_notifier);
711 
712     g_free(nodename);
713 }
714 
715 static void create_virtio_devices(const VirtMachineState *vms, qemu_irq *pic)
716 {
717     int i;
718     hwaddr size = vms->memmap[VIRT_MMIO].size;
719 
720     /* We create the transports in forwards order. Since qbus_realize()
721      * prepends (not appends) new child buses, the incrementing loop below will
722      * create a list of virtio-mmio buses with decreasing base addresses.
723      *
724      * When a -device option is processed from the command line,
725      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
726      * order. The upshot is that -device options in increasing command line
727      * order are mapped to virtio-mmio buses with decreasing base addresses.
728      *
729      * When this code was originally written, that arrangement ensured that the
730      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
731      * the first -device on the command line. (The end-to-end order is a
732      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
733      * guest kernel's name-to-address assignment strategy.)
734      *
735      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
736      * the message, if not necessarily the code, of commit 70161ff336.
737      * Therefore the loop now establishes the inverse of the original intent.
738      *
739      * Unfortunately, we can't counteract the kernel change by reversing the
740      * loop; it would break existing command lines.
741      *
742      * In any case, the kernel makes no guarantee about the stability of
743      * enumeration order of virtio devices (as demonstrated by it changing
744      * between kernel versions). For reliable and stable identification
745      * of disks users must use UUIDs or similar mechanisms.
746      */
747     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
748         int irq = vms->irqmap[VIRT_MMIO] + i;
749         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
750 
751         sysbus_create_simple("virtio-mmio", base, pic[irq]);
752     }
753 
754     /* We add dtb nodes in reverse order so that they appear in the finished
755      * device tree lowest address first.
756      *
757      * Note that this mapping is independent of the loop above. The previous
758      * loop influences virtio device to virtio transport assignment, whereas
759      * this loop controls how virtio transports are laid out in the dtb.
760      */
761     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
762         char *nodename;
763         int irq = vms->irqmap[VIRT_MMIO] + i;
764         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
765 
766         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
767         qemu_fdt_add_subnode(vms->fdt, nodename);
768         qemu_fdt_setprop_string(vms->fdt, nodename,
769                                 "compatible", "virtio,mmio");
770         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
771                                      2, base, 2, size);
772         qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
773                                GIC_FDT_IRQ_TYPE_SPI, irq,
774                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
775         qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
776         g_free(nodename);
777     }
778 }
779 
780 static void create_one_flash(const char *name, hwaddr flashbase,
781                              hwaddr flashsize, const char *file,
782                              MemoryRegion *sysmem)
783 {
784     /* Create and map a single flash device. We use the same
785      * parameters as the flash devices on the Versatile Express board.
786      */
787     DriveInfo *dinfo = drive_get_next(IF_PFLASH);
788     DeviceState *dev = qdev_create(NULL, "cfi.pflash01");
789     SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
790     const uint64_t sectorlength = 256 * 1024;
791 
792     if (dinfo) {
793         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
794                             &error_abort);
795     }
796 
797     qdev_prop_set_uint32(dev, "num-blocks", flashsize / sectorlength);
798     qdev_prop_set_uint64(dev, "sector-length", sectorlength);
799     qdev_prop_set_uint8(dev, "width", 4);
800     qdev_prop_set_uint8(dev, "device-width", 2);
801     qdev_prop_set_bit(dev, "big-endian", false);
802     qdev_prop_set_uint16(dev, "id0", 0x89);
803     qdev_prop_set_uint16(dev, "id1", 0x18);
804     qdev_prop_set_uint16(dev, "id2", 0x00);
805     qdev_prop_set_uint16(dev, "id3", 0x00);
806     qdev_prop_set_string(dev, "name", name);
807     qdev_init_nofail(dev);
808 
809     memory_region_add_subregion(sysmem, flashbase,
810                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0));
811 
812     if (file) {
813         char *fn;
814         int image_size;
815 
816         if (drive_get(IF_PFLASH, 0, 0)) {
817             error_report("The contents of the first flash device may be "
818                          "specified with -bios or with -drive if=pflash... "
819                          "but you cannot use both options at once");
820             exit(1);
821         }
822         fn = qemu_find_file(QEMU_FILE_TYPE_BIOS, file);
823         if (!fn) {
824             error_report("Could not find ROM image '%s'", file);
825             exit(1);
826         }
827         image_size = load_image_mr(fn, sysbus_mmio_get_region(sbd, 0));
828         g_free(fn);
829         if (image_size < 0) {
830             error_report("Could not load ROM image '%s'", file);
831             exit(1);
832         }
833     }
834 }
835 
836 static void create_flash(const VirtMachineState *vms,
837                          MemoryRegion *sysmem,
838                          MemoryRegion *secure_sysmem)
839 {
840     /* Create two flash devices to fill the VIRT_FLASH space in the memmap.
841      * Any file passed via -bios goes in the first of these.
842      * sysmem is the system memory space. secure_sysmem is the secure view
843      * of the system, and the first flash device should be made visible only
844      * there. The second flash device is visible to both secure and nonsecure.
845      * If sysmem == secure_sysmem this means there is no separate Secure
846      * address space and both flash devices are generally visible.
847      */
848     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
849     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
850     char *nodename;
851 
852     create_one_flash("virt.flash0", flashbase, flashsize,
853                      bios_name, secure_sysmem);
854     create_one_flash("virt.flash1", flashbase + flashsize, flashsize,
855                      NULL, sysmem);
856 
857     if (sysmem == secure_sysmem) {
858         /* Report both flash devices as a single node in the DT */
859         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
860         qemu_fdt_add_subnode(vms->fdt, nodename);
861         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
862         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
863                                      2, flashbase, 2, flashsize,
864                                      2, flashbase + flashsize, 2, flashsize);
865         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
866         g_free(nodename);
867     } else {
868         /* Report the devices as separate nodes so we can mark one as
869          * only visible to the secure world.
870          */
871         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
872         qemu_fdt_add_subnode(vms->fdt, nodename);
873         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
874         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
875                                      2, flashbase, 2, flashsize);
876         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
877         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
878         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
879         g_free(nodename);
880 
881         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
882         qemu_fdt_add_subnode(vms->fdt, nodename);
883         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
884         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
885                                      2, flashbase + flashsize, 2, flashsize);
886         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
887         g_free(nodename);
888     }
889 }
890 
891 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
892 {
893     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
894     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
895     FWCfgState *fw_cfg;
896     char *nodename;
897 
898     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
899     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)smp_cpus);
900 
901     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
902     qemu_fdt_add_subnode(vms->fdt, nodename);
903     qemu_fdt_setprop_string(vms->fdt, nodename,
904                             "compatible", "qemu,fw-cfg-mmio");
905     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
906                                  2, base, 2, size);
907     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
908     g_free(nodename);
909     return fw_cfg;
910 }
911 
912 static void create_pcie_irq_map(const VirtMachineState *vms,
913                                 uint32_t gic_phandle,
914                                 int first_irq, const char *nodename)
915 {
916     int devfn, pin;
917     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
918     uint32_t *irq_map = full_irq_map;
919 
920     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
921         for (pin = 0; pin < 4; pin++) {
922             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
923             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
924             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
925             int i;
926 
927             uint32_t map[] = {
928                 devfn << 8, 0, 0,                           /* devfn */
929                 pin + 1,                                    /* PCI pin */
930                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
931 
932             /* Convert map to big endian */
933             for (i = 0; i < 10; i++) {
934                 irq_map[i] = cpu_to_be32(map[i]);
935             }
936             irq_map += 10;
937         }
938     }
939 
940     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
941                      full_irq_map, sizeof(full_irq_map));
942 
943     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
944                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
945                            0x7           /* PCI irq */);
946 }
947 
948 static void create_smmu(const VirtMachineState *vms, qemu_irq *pic,
949                         PCIBus *bus)
950 {
951     char *node;
952     const char compat[] = "arm,smmu-v3";
953     int irq =  vms->irqmap[VIRT_SMMU];
954     int i;
955     hwaddr base = vms->memmap[VIRT_SMMU].base;
956     hwaddr size = vms->memmap[VIRT_SMMU].size;
957     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
958     DeviceState *dev;
959 
960     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
961         return;
962     }
963 
964     dev = qdev_create(NULL, "arm-smmuv3");
965 
966     object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
967                              &error_abort);
968     qdev_init_nofail(dev);
969     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
970     for (i = 0; i < NUM_SMMU_IRQS; i++) {
971         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
972     }
973 
974     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
975     qemu_fdt_add_subnode(vms->fdt, node);
976     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
977     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
978 
979     qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
980             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
981             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
982             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
983             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
984 
985     qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
986                      sizeof(irq_names));
987 
988     qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
989     qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
990     qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
991 
992     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
993 
994     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
995     g_free(node);
996 }
997 
998 static void create_pcie(VirtMachineState *vms, qemu_irq *pic)
999 {
1000     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1001     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1002     hwaddr base_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].base;
1003     hwaddr size_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].size;
1004     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1005     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1006     hwaddr base_ecam = vms->memmap[VIRT_PCIE_ECAM].base;
1007     hwaddr size_ecam = vms->memmap[VIRT_PCIE_ECAM].size;
1008     hwaddr base = base_mmio;
1009     int nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1010     int irq = vms->irqmap[VIRT_PCIE];
1011     MemoryRegion *mmio_alias;
1012     MemoryRegion *mmio_reg;
1013     MemoryRegion *ecam_alias;
1014     MemoryRegion *ecam_reg;
1015     DeviceState *dev;
1016     char *nodename;
1017     int i;
1018     PCIHostState *pci;
1019 
1020     dev = qdev_create(NULL, TYPE_GPEX_HOST);
1021     qdev_init_nofail(dev);
1022 
1023     /* Map only the first size_ecam bytes of ECAM space */
1024     ecam_alias = g_new0(MemoryRegion, 1);
1025     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1026     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1027                              ecam_reg, 0, size_ecam);
1028     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1029 
1030     /* Map the MMIO window into system address space so as to expose
1031      * the section of PCI MMIO space which starts at the same base address
1032      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1033      * the window).
1034      */
1035     mmio_alias = g_new0(MemoryRegion, 1);
1036     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1037     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1038                              mmio_reg, base_mmio, size_mmio);
1039     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1040 
1041     if (vms->highmem) {
1042         /* Map high MMIO space */
1043         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1044 
1045         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1046                                  mmio_reg, base_mmio_high, size_mmio_high);
1047         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1048                                     high_mmio_alias);
1049     }
1050 
1051     /* Map IO port space */
1052     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1053 
1054     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1055         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
1056         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1057     }
1058 
1059     pci = PCI_HOST_BRIDGE(dev);
1060     if (pci->bus) {
1061         for (i = 0; i < nb_nics; i++) {
1062             NICInfo *nd = &nd_table[i];
1063 
1064             if (!nd->model) {
1065                 nd->model = g_strdup("virtio");
1066             }
1067 
1068             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1069         }
1070     }
1071 
1072     nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1073     qemu_fdt_add_subnode(vms->fdt, nodename);
1074     qemu_fdt_setprop_string(vms->fdt, nodename,
1075                             "compatible", "pci-host-ecam-generic");
1076     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1077     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1078     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1079     qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1080     qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1081                            nr_pcie_buses - 1);
1082     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1083 
1084     if (vms->msi_phandle) {
1085         qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1086                                vms->msi_phandle);
1087     }
1088 
1089     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1090                                  2, base_ecam, 2, size_ecam);
1091 
1092     if (vms->highmem) {
1093         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1094                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1095                                      2, base_pio, 2, size_pio,
1096                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1097                                      2, base_mmio, 2, size_mmio,
1098                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1099                                      2, base_mmio_high,
1100                                      2, base_mmio_high, 2, size_mmio_high);
1101     } else {
1102         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1103                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1104                                      2, base_pio, 2, size_pio,
1105                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1106                                      2, base_mmio, 2, size_mmio);
1107     }
1108 
1109     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1110     create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1111 
1112     if (vms->iommu) {
1113         vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1114 
1115         create_smmu(vms, pic, pci->bus);
1116 
1117         qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1118                                0x0, vms->iommu_phandle, 0x0, 0x10000);
1119     }
1120 
1121     g_free(nodename);
1122 }
1123 
1124 static void create_platform_bus(VirtMachineState *vms, qemu_irq *pic)
1125 {
1126     DeviceState *dev;
1127     SysBusDevice *s;
1128     int i;
1129     ARMPlatformBusFDTParams *fdt_params = g_new(ARMPlatformBusFDTParams, 1);
1130     MemoryRegion *sysmem = get_system_memory();
1131 
1132     platform_bus_params.platform_bus_base = vms->memmap[VIRT_PLATFORM_BUS].base;
1133     platform_bus_params.platform_bus_size = vms->memmap[VIRT_PLATFORM_BUS].size;
1134     platform_bus_params.platform_bus_first_irq = vms->irqmap[VIRT_PLATFORM_BUS];
1135     platform_bus_params.platform_bus_num_irqs = PLATFORM_BUS_NUM_IRQS;
1136 
1137     fdt_params->system_params = &platform_bus_params;
1138     fdt_params->binfo = &vms->bootinfo;
1139     fdt_params->intc = "/intc";
1140     /*
1141      * register a machine init done notifier that creates the device tree
1142      * nodes of the platform bus and its children dynamic sysbus devices
1143      */
1144     arm_register_platform_bus_fdt_creator(fdt_params);
1145 
1146     dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1147     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1148     qdev_prop_set_uint32(dev, "num_irqs",
1149         platform_bus_params.platform_bus_num_irqs);
1150     qdev_prop_set_uint32(dev, "mmio_size",
1151         platform_bus_params.platform_bus_size);
1152     qdev_init_nofail(dev);
1153     vms->platform_bus_dev = dev;
1154     s = SYS_BUS_DEVICE(dev);
1155 
1156     for (i = 0; i < platform_bus_params.platform_bus_num_irqs; i++) {
1157         int irqn = platform_bus_params.platform_bus_first_irq + i;
1158         sysbus_connect_irq(s, i, pic[irqn]);
1159     }
1160 
1161     memory_region_add_subregion(sysmem,
1162                                 platform_bus_params.platform_bus_base,
1163                                 sysbus_mmio_get_region(s, 0));
1164 }
1165 
1166 static void create_secure_ram(VirtMachineState *vms,
1167                               MemoryRegion *secure_sysmem)
1168 {
1169     MemoryRegion *secram = g_new(MemoryRegion, 1);
1170     char *nodename;
1171     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1172     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1173 
1174     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1175                            &error_fatal);
1176     memory_region_add_subregion(secure_sysmem, base, secram);
1177 
1178     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1179     qemu_fdt_add_subnode(vms->fdt, nodename);
1180     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1181     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1182     qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1183     qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1184 
1185     g_free(nodename);
1186 }
1187 
1188 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1189 {
1190     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1191                                                  bootinfo);
1192 
1193     *fdt_size = board->fdt_size;
1194     return board->fdt;
1195 }
1196 
1197 static void virt_build_smbios(VirtMachineState *vms)
1198 {
1199     MachineClass *mc = MACHINE_GET_CLASS(vms);
1200     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1201     uint8_t *smbios_tables, *smbios_anchor;
1202     size_t smbios_tables_len, smbios_anchor_len;
1203     const char *product = "QEMU Virtual Machine";
1204 
1205     if (!vms->fw_cfg) {
1206         return;
1207     }
1208 
1209     if (kvm_enabled()) {
1210         product = "KVM Virtual Machine";
1211     }
1212 
1213     smbios_set_defaults("QEMU", product,
1214                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1215                         true, SMBIOS_ENTRY_POINT_30);
1216 
1217     smbios_get_tables(NULL, 0, &smbios_tables, &smbios_tables_len,
1218                       &smbios_anchor, &smbios_anchor_len);
1219 
1220     if (smbios_anchor) {
1221         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1222                         smbios_tables, smbios_tables_len);
1223         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1224                         smbios_anchor, smbios_anchor_len);
1225     }
1226 }
1227 
1228 static
1229 void virt_machine_done(Notifier *notifier, void *data)
1230 {
1231     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1232                                          machine_done);
1233 
1234     virt_acpi_setup(vms);
1235     virt_build_smbios(vms);
1236 }
1237 
1238 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1239 {
1240     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1241     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1242 
1243     if (!vmc->disallow_affinity_adjustment) {
1244         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1245          * GIC's target-list limitations. 32-bit KVM hosts currently
1246          * always create clusters of 4 CPUs, but that is expected to
1247          * change when they gain support for gicv3. When KVM is enabled
1248          * it will override the changes we make here, therefore our
1249          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1250          * and to improve SGI efficiency.
1251          */
1252         if (vms->gic_version == 3) {
1253             clustersz = GICV3_TARGETLIST_BITS;
1254         } else {
1255             clustersz = GIC_TARGETLIST_BITS;
1256         }
1257     }
1258     return arm_cpu_mp_affinity(idx, clustersz);
1259 }
1260 
1261 static void machvirt_init(MachineState *machine)
1262 {
1263     VirtMachineState *vms = VIRT_MACHINE(machine);
1264     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1265     MachineClass *mc = MACHINE_GET_CLASS(machine);
1266     const CPUArchIdList *possible_cpus;
1267     qemu_irq pic[NUM_IRQS];
1268     MemoryRegion *sysmem = get_system_memory();
1269     MemoryRegion *secure_sysmem = NULL;
1270     int n, virt_max_cpus;
1271     MemoryRegion *ram = g_new(MemoryRegion, 1);
1272     bool firmware_loaded = bios_name || drive_get(IF_PFLASH, 0, 0);
1273 
1274     /* We can probe only here because during property set
1275      * KVM is not available yet
1276      */
1277     if (vms->gic_version <= 0) {
1278         /* "host" or "max" */
1279         if (!kvm_enabled()) {
1280             if (vms->gic_version == 0) {
1281                 error_report("gic-version=host requires KVM");
1282                 exit(1);
1283             } else {
1284                 /* "max": currently means 3 for TCG */
1285                 vms->gic_version = 3;
1286             }
1287         } else {
1288             vms->gic_version = kvm_arm_vgic_probe();
1289             if (!vms->gic_version) {
1290                 error_report(
1291                     "Unable to determine GIC version supported by host");
1292                 exit(1);
1293             }
1294         }
1295     }
1296 
1297     if (!cpu_type_valid(machine->cpu_type)) {
1298         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1299         exit(1);
1300     }
1301 
1302     /* If we have an EL3 boot ROM then the assumption is that it will
1303      * implement PSCI itself, so disable QEMU's internal implementation
1304      * so it doesn't get in the way. Instead of starting secondary
1305      * CPUs in PSCI powerdown state we will start them all running and
1306      * let the boot ROM sort them out.
1307      * The usual case is that we do use QEMU's PSCI implementation;
1308      * if the guest has EL2 then we will use SMC as the conduit,
1309      * and otherwise we will use HVC (for backwards compatibility and
1310      * because if we're using KVM then we must use HVC).
1311      */
1312     if (vms->secure && firmware_loaded) {
1313         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1314     } else if (vms->virt) {
1315         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1316     } else {
1317         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1318     }
1319 
1320     /* The maximum number of CPUs depends on the GIC version, or on how
1321      * many redistributors we can fit into the memory map.
1322      */
1323     if (vms->gic_version == 3) {
1324         virt_max_cpus = vms->memmap[VIRT_GIC_REDIST].size / 0x20000;
1325     } else {
1326         virt_max_cpus = GIC_NCPU;
1327     }
1328 
1329     if (max_cpus > virt_max_cpus) {
1330         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1331                      "supported by machine 'mach-virt' (%d)",
1332                      max_cpus, virt_max_cpus);
1333         exit(1);
1334     }
1335 
1336     vms->smp_cpus = smp_cpus;
1337 
1338     if (machine->ram_size > vms->memmap[VIRT_MEM].size) {
1339         error_report("mach-virt: cannot model more than %dGB RAM", RAMLIMIT_GB);
1340         exit(1);
1341     }
1342 
1343     if (vms->virt && kvm_enabled()) {
1344         error_report("mach-virt: KVM does not support providing "
1345                      "Virtualization extensions to the guest CPU");
1346         exit(1);
1347     }
1348 
1349     if (vms->secure) {
1350         if (kvm_enabled()) {
1351             error_report("mach-virt: KVM does not support Security extensions");
1352             exit(1);
1353         }
1354 
1355         /* The Secure view of the world is the same as the NonSecure,
1356          * but with a few extra devices. Create it as a container region
1357          * containing the system memory at low priority; any secure-only
1358          * devices go in at higher priority and take precedence.
1359          */
1360         secure_sysmem = g_new(MemoryRegion, 1);
1361         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1362                            UINT64_MAX);
1363         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1364     }
1365 
1366     create_fdt(vms);
1367 
1368     possible_cpus = mc->possible_cpu_arch_ids(machine);
1369     for (n = 0; n < possible_cpus->len; n++) {
1370         Object *cpuobj;
1371         CPUState *cs;
1372 
1373         if (n >= smp_cpus) {
1374             break;
1375         }
1376 
1377         cpuobj = object_new(possible_cpus->cpus[n].type);
1378         object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
1379                                 "mp-affinity", NULL);
1380 
1381         cs = CPU(cpuobj);
1382         cs->cpu_index = n;
1383 
1384         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1385                           &error_fatal);
1386 
1387         if (!vms->secure) {
1388             object_property_set_bool(cpuobj, false, "has_el3", NULL);
1389         }
1390 
1391         if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
1392             object_property_set_bool(cpuobj, false, "has_el2", NULL);
1393         }
1394 
1395         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1396             object_property_set_int(cpuobj, vms->psci_conduit,
1397                                     "psci-conduit", NULL);
1398 
1399             /* Secondary CPUs start in PSCI powered-down state */
1400             if (n > 0) {
1401                 object_property_set_bool(cpuobj, true,
1402                                          "start-powered-off", NULL);
1403             }
1404         }
1405 
1406         if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1407             object_property_set_bool(cpuobj, false, "pmu", NULL);
1408         }
1409 
1410         if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1411             object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
1412                                     "reset-cbar", &error_abort);
1413         }
1414 
1415         object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1416                                  &error_abort);
1417         if (vms->secure) {
1418             object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1419                                      "secure-memory", &error_abort);
1420         }
1421 
1422         object_property_set_bool(cpuobj, true, "realized", &error_fatal);
1423         object_unref(cpuobj);
1424     }
1425     fdt_add_timer_nodes(vms);
1426     fdt_add_cpu_nodes(vms);
1427 
1428     memory_region_allocate_system_memory(ram, NULL, "mach-virt.ram",
1429                                          machine->ram_size);
1430     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base, ram);
1431 
1432     create_flash(vms, sysmem, secure_sysmem ? secure_sysmem : sysmem);
1433 
1434     create_gic(vms, pic);
1435 
1436     fdt_add_pmu_nodes(vms);
1437 
1438     create_uart(vms, pic, VIRT_UART, sysmem, serial_hd(0));
1439 
1440     if (vms->secure) {
1441         create_secure_ram(vms, secure_sysmem);
1442         create_uart(vms, pic, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
1443     }
1444 
1445     create_rtc(vms, pic);
1446 
1447     create_pcie(vms, pic);
1448 
1449     create_gpio(vms, pic);
1450 
1451     /* Create mmio transports, so the user can create virtio backends
1452      * (which will be automatically plugged in to the transports). If
1453      * no backend is created the transport will just sit harmlessly idle.
1454      */
1455     create_virtio_devices(vms, pic);
1456 
1457     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1458     rom_set_fw(vms->fw_cfg);
1459 
1460     vms->machine_done.notify = virt_machine_done;
1461     qemu_add_machine_init_done_notifier(&vms->machine_done);
1462 
1463     vms->bootinfo.ram_size = machine->ram_size;
1464     vms->bootinfo.kernel_filename = machine->kernel_filename;
1465     vms->bootinfo.kernel_cmdline = machine->kernel_cmdline;
1466     vms->bootinfo.initrd_filename = machine->initrd_filename;
1467     vms->bootinfo.nb_cpus = smp_cpus;
1468     vms->bootinfo.board_id = -1;
1469     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1470     vms->bootinfo.get_dtb = machvirt_dtb;
1471     vms->bootinfo.firmware_loaded = firmware_loaded;
1472     arm_load_kernel(ARM_CPU(first_cpu), &vms->bootinfo);
1473 
1474     /*
1475      * arm_load_kernel machine init done notifier registration must
1476      * happen before the platform_bus_create call. In this latter,
1477      * another notifier is registered which adds platform bus nodes.
1478      * Notifiers are executed in registration reverse order.
1479      */
1480     create_platform_bus(vms, pic);
1481 }
1482 
1483 static bool virt_get_secure(Object *obj, Error **errp)
1484 {
1485     VirtMachineState *vms = VIRT_MACHINE(obj);
1486 
1487     return vms->secure;
1488 }
1489 
1490 static void virt_set_secure(Object *obj, bool value, Error **errp)
1491 {
1492     VirtMachineState *vms = VIRT_MACHINE(obj);
1493 
1494     vms->secure = value;
1495 }
1496 
1497 static bool virt_get_virt(Object *obj, Error **errp)
1498 {
1499     VirtMachineState *vms = VIRT_MACHINE(obj);
1500 
1501     return vms->virt;
1502 }
1503 
1504 static void virt_set_virt(Object *obj, bool value, Error **errp)
1505 {
1506     VirtMachineState *vms = VIRT_MACHINE(obj);
1507 
1508     vms->virt = value;
1509 }
1510 
1511 static bool virt_get_highmem(Object *obj, Error **errp)
1512 {
1513     VirtMachineState *vms = VIRT_MACHINE(obj);
1514 
1515     return vms->highmem;
1516 }
1517 
1518 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1519 {
1520     VirtMachineState *vms = VIRT_MACHINE(obj);
1521 
1522     vms->highmem = value;
1523 }
1524 
1525 static bool virt_get_its(Object *obj, Error **errp)
1526 {
1527     VirtMachineState *vms = VIRT_MACHINE(obj);
1528 
1529     return vms->its;
1530 }
1531 
1532 static void virt_set_its(Object *obj, bool value, Error **errp)
1533 {
1534     VirtMachineState *vms = VIRT_MACHINE(obj);
1535 
1536     vms->its = value;
1537 }
1538 
1539 static char *virt_get_gic_version(Object *obj, Error **errp)
1540 {
1541     VirtMachineState *vms = VIRT_MACHINE(obj);
1542     const char *val = vms->gic_version == 3 ? "3" : "2";
1543 
1544     return g_strdup(val);
1545 }
1546 
1547 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1548 {
1549     VirtMachineState *vms = VIRT_MACHINE(obj);
1550 
1551     if (!strcmp(value, "3")) {
1552         vms->gic_version = 3;
1553     } else if (!strcmp(value, "2")) {
1554         vms->gic_version = 2;
1555     } else if (!strcmp(value, "host")) {
1556         vms->gic_version = 0; /* Will probe later */
1557     } else if (!strcmp(value, "max")) {
1558         vms->gic_version = -1; /* Will probe later */
1559     } else {
1560         error_setg(errp, "Invalid gic-version value");
1561         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
1562     }
1563 }
1564 
1565 static char *virt_get_iommu(Object *obj, Error **errp)
1566 {
1567     VirtMachineState *vms = VIRT_MACHINE(obj);
1568 
1569     switch (vms->iommu) {
1570     case VIRT_IOMMU_NONE:
1571         return g_strdup("none");
1572     case VIRT_IOMMU_SMMUV3:
1573         return g_strdup("smmuv3");
1574     default:
1575         g_assert_not_reached();
1576     }
1577 }
1578 
1579 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
1580 {
1581     VirtMachineState *vms = VIRT_MACHINE(obj);
1582 
1583     if (!strcmp(value, "smmuv3")) {
1584         vms->iommu = VIRT_IOMMU_SMMUV3;
1585     } else if (!strcmp(value, "none")) {
1586         vms->iommu = VIRT_IOMMU_NONE;
1587     } else {
1588         error_setg(errp, "Invalid iommu value");
1589         error_append_hint(errp, "Valid values are none, smmuv3.\n");
1590     }
1591 }
1592 
1593 static CpuInstanceProperties
1594 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
1595 {
1596     MachineClass *mc = MACHINE_GET_CLASS(ms);
1597     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1598 
1599     assert(cpu_index < possible_cpus->len);
1600     return possible_cpus->cpus[cpu_index].props;
1601 }
1602 
1603 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
1604 {
1605     return idx % nb_numa_nodes;
1606 }
1607 
1608 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
1609 {
1610     int n;
1611     VirtMachineState *vms = VIRT_MACHINE(ms);
1612 
1613     if (ms->possible_cpus) {
1614         assert(ms->possible_cpus->len == max_cpus);
1615         return ms->possible_cpus;
1616     }
1617 
1618     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1619                                   sizeof(CPUArchId) * max_cpus);
1620     ms->possible_cpus->len = max_cpus;
1621     for (n = 0; n < ms->possible_cpus->len; n++) {
1622         ms->possible_cpus->cpus[n].type = ms->cpu_type;
1623         ms->possible_cpus->cpus[n].arch_id =
1624             virt_cpu_mp_affinity(vms, n);
1625         ms->possible_cpus->cpus[n].props.has_thread_id = true;
1626         ms->possible_cpus->cpus[n].props.thread_id = n;
1627     }
1628     return ms->possible_cpus;
1629 }
1630 
1631 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1632                                         DeviceState *dev, Error **errp)
1633 {
1634     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1635 
1636     if (vms->platform_bus_dev) {
1637         if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
1638             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
1639                                      SYS_BUS_DEVICE(dev));
1640         }
1641     }
1642 }
1643 
1644 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
1645                                                         DeviceState *dev)
1646 {
1647     if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
1648         return HOTPLUG_HANDLER(machine);
1649     }
1650 
1651     return NULL;
1652 }
1653 
1654 static void virt_machine_class_init(ObjectClass *oc, void *data)
1655 {
1656     MachineClass *mc = MACHINE_CLASS(oc);
1657     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1658 
1659     mc->init = machvirt_init;
1660     /* Start max_cpus at the maximum QEMU supports. We'll further restrict
1661      * it later in machvirt_init, where we have more information about the
1662      * configuration of the particular instance.
1663      */
1664     mc->max_cpus = 255;
1665     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
1666     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
1667     mc->block_default_type = IF_VIRTIO;
1668     mc->no_cdrom = 1;
1669     mc->pci_allow_0_address = true;
1670     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
1671     mc->minimum_page_bits = 12;
1672     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
1673     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
1674     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
1675     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
1676     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
1677     hc->plug = virt_machine_device_plug_cb;
1678 }
1679 
1680 static const TypeInfo virt_machine_info = {
1681     .name          = TYPE_VIRT_MACHINE,
1682     .parent        = TYPE_MACHINE,
1683     .abstract      = true,
1684     .instance_size = sizeof(VirtMachineState),
1685     .class_size    = sizeof(VirtMachineClass),
1686     .class_init    = virt_machine_class_init,
1687     .interfaces = (InterfaceInfo[]) {
1688          { TYPE_HOTPLUG_HANDLER },
1689          { }
1690     },
1691 };
1692 
1693 static void machvirt_machine_init(void)
1694 {
1695     type_register_static(&virt_machine_info);
1696 }
1697 type_init(machvirt_machine_init);
1698 
1699 static void virt_2_12_instance_init(Object *obj)
1700 {
1701     VirtMachineState *vms = VIRT_MACHINE(obj);
1702     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1703 
1704     /* EL3 is disabled by default on virt: this makes us consistent
1705      * between KVM and TCG for this board, and it also allows us to
1706      * boot UEFI blobs which assume no TrustZone support.
1707      */
1708     vms->secure = false;
1709     object_property_add_bool(obj, "secure", virt_get_secure,
1710                              virt_set_secure, NULL);
1711     object_property_set_description(obj, "secure",
1712                                     "Set on/off to enable/disable the ARM "
1713                                     "Security Extensions (TrustZone)",
1714                                     NULL);
1715 
1716     /* EL2 is also disabled by default, for similar reasons */
1717     vms->virt = false;
1718     object_property_add_bool(obj, "virtualization", virt_get_virt,
1719                              virt_set_virt, NULL);
1720     object_property_set_description(obj, "virtualization",
1721                                     "Set on/off to enable/disable emulating a "
1722                                     "guest CPU which implements the ARM "
1723                                     "Virtualization Extensions",
1724                                     NULL);
1725 
1726     /* High memory is enabled by default */
1727     vms->highmem = true;
1728     object_property_add_bool(obj, "highmem", virt_get_highmem,
1729                              virt_set_highmem, NULL);
1730     object_property_set_description(obj, "highmem",
1731                                     "Set on/off to enable/disable using "
1732                                     "physical address space above 32 bits",
1733                                     NULL);
1734     /* Default GIC type is v2 */
1735     vms->gic_version = 2;
1736     object_property_add_str(obj, "gic-version", virt_get_gic_version,
1737                         virt_set_gic_version, NULL);
1738     object_property_set_description(obj, "gic-version",
1739                                     "Set GIC version. "
1740                                     "Valid values are 2, 3 and host", NULL);
1741 
1742     if (vmc->no_its) {
1743         vms->its = false;
1744     } else {
1745         /* Default allows ITS instantiation */
1746         vms->its = true;
1747         object_property_add_bool(obj, "its", virt_get_its,
1748                                  virt_set_its, NULL);
1749         object_property_set_description(obj, "its",
1750                                         "Set on/off to enable/disable "
1751                                         "ITS instantiation",
1752                                         NULL);
1753     }
1754 
1755     /* Default disallows iommu instantiation */
1756     vms->iommu = VIRT_IOMMU_NONE;
1757     object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu, NULL);
1758     object_property_set_description(obj, "iommu",
1759                                     "Set the IOMMU type. "
1760                                     "Valid values are none and smmuv3",
1761                                     NULL);
1762 
1763     vms->memmap = a15memmap;
1764     vms->irqmap = a15irqmap;
1765 }
1766 
1767 static void virt_machine_2_12_options(MachineClass *mc)
1768 {
1769 }
1770 DEFINE_VIRT_MACHINE_AS_LATEST(2, 12)
1771 
1772 #define VIRT_COMPAT_2_11 \
1773     HW_COMPAT_2_11
1774 
1775 static void virt_2_11_instance_init(Object *obj)
1776 {
1777     virt_2_12_instance_init(obj);
1778 }
1779 
1780 static void virt_machine_2_11_options(MachineClass *mc)
1781 {
1782     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1783 
1784     virt_machine_2_12_options(mc);
1785     SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_11);
1786     vmc->smbios_old_sys_ver = true;
1787 }
1788 DEFINE_VIRT_MACHINE(2, 11)
1789 
1790 #define VIRT_COMPAT_2_10 \
1791     HW_COMPAT_2_10
1792 
1793 static void virt_2_10_instance_init(Object *obj)
1794 {
1795     virt_2_11_instance_init(obj);
1796 }
1797 
1798 static void virt_machine_2_10_options(MachineClass *mc)
1799 {
1800     virt_machine_2_11_options(mc);
1801     SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_10);
1802 }
1803 DEFINE_VIRT_MACHINE(2, 10)
1804 
1805 #define VIRT_COMPAT_2_9 \
1806     HW_COMPAT_2_9
1807 
1808 static void virt_2_9_instance_init(Object *obj)
1809 {
1810     virt_2_10_instance_init(obj);
1811 }
1812 
1813 static void virt_machine_2_9_options(MachineClass *mc)
1814 {
1815     virt_machine_2_10_options(mc);
1816     SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_9);
1817 }
1818 DEFINE_VIRT_MACHINE(2, 9)
1819 
1820 #define VIRT_COMPAT_2_8 \
1821     HW_COMPAT_2_8
1822 
1823 static void virt_2_8_instance_init(Object *obj)
1824 {
1825     virt_2_9_instance_init(obj);
1826 }
1827 
1828 static void virt_machine_2_8_options(MachineClass *mc)
1829 {
1830     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1831 
1832     virt_machine_2_9_options(mc);
1833     SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_8);
1834     /* For 2.8 and earlier we falsely claimed in the DT that
1835      * our timers were edge-triggered, not level-triggered.
1836      */
1837     vmc->claim_edge_triggered_timers = true;
1838 }
1839 DEFINE_VIRT_MACHINE(2, 8)
1840 
1841 #define VIRT_COMPAT_2_7 \
1842     HW_COMPAT_2_7
1843 
1844 static void virt_2_7_instance_init(Object *obj)
1845 {
1846     virt_2_8_instance_init(obj);
1847 }
1848 
1849 static void virt_machine_2_7_options(MachineClass *mc)
1850 {
1851     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1852 
1853     virt_machine_2_8_options(mc);
1854     SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_7);
1855     /* ITS was introduced with 2.8 */
1856     vmc->no_its = true;
1857     /* Stick with 1K pages for migration compatibility */
1858     mc->minimum_page_bits = 0;
1859 }
1860 DEFINE_VIRT_MACHINE(2, 7)
1861 
1862 #define VIRT_COMPAT_2_6 \
1863     HW_COMPAT_2_6
1864 
1865 static void virt_2_6_instance_init(Object *obj)
1866 {
1867     virt_2_7_instance_init(obj);
1868 }
1869 
1870 static void virt_machine_2_6_options(MachineClass *mc)
1871 {
1872     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1873 
1874     virt_machine_2_7_options(mc);
1875     SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_6);
1876     vmc->disallow_affinity_adjustment = true;
1877     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
1878     vmc->no_pmu = true;
1879 }
1880 DEFINE_VIRT_MACHINE(2, 6)
1881