xref: /openbmc/qemu/hw/arm/virt.c (revision 47451466)
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/device_tree.h"
42 #include "sysemu/numa.h"
43 #include "sysemu/sysemu.h"
44 #include "sysemu/kvm.h"
45 #include "hw/compat.h"
46 #include "hw/loader.h"
47 #include "exec/address-spaces.h"
48 #include "qemu/bitops.h"
49 #include "qemu/error-report.h"
50 #include "hw/pci-host/gpex.h"
51 #include "hw/arm/sysbus-fdt.h"
52 #include "hw/platform-bus.h"
53 #include "hw/arm/fdt.h"
54 #include "hw/intc/arm_gic.h"
55 #include "hw/intc/arm_gicv3_common.h"
56 #include "kvm_arm.h"
57 #include "hw/smbios/smbios.h"
58 #include "qapi/visitor.h"
59 #include "standard-headers/linux/input.h"
60 #include "hw/arm/smmuv3.h"
61 
62 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
63     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
64                                                     void *data) \
65     { \
66         MachineClass *mc = MACHINE_CLASS(oc); \
67         virt_machine_##major##_##minor##_options(mc); \
68         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
69         if (latest) { \
70             mc->alias = "virt"; \
71         } \
72     } \
73     static const TypeInfo machvirt_##major##_##minor##_info = { \
74         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
75         .parent = TYPE_VIRT_MACHINE, \
76         .instance_init = virt_##major##_##minor##_instance_init, \
77         .class_init = virt_##major##_##minor##_class_init, \
78     }; \
79     static void machvirt_machine_##major##_##minor##_init(void) \
80     { \
81         type_register_static(&machvirt_##major##_##minor##_info); \
82     } \
83     type_init(machvirt_machine_##major##_##minor##_init);
84 
85 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
86     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
87 #define DEFINE_VIRT_MACHINE(major, minor) \
88     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
89 
90 
91 /* Number of external interrupt lines to configure the GIC with */
92 #define NUM_IRQS 256
93 
94 #define PLATFORM_BUS_NUM_IRQS 64
95 
96 /* RAM limit in GB. Since VIRT_MEM starts at the 1GB mark, this means
97  * RAM can go up to the 256GB mark, leaving 256GB of the physical
98  * address space unallocated and free for future use between 256G and 512G.
99  * If we need to provide more RAM to VMs in the future then we need to:
100  *  * allocate a second bank of RAM starting at 2TB and working up
101  *  * fix the DT and ACPI table generation code in QEMU to correctly
102  *    report two split lumps of RAM to the guest
103  *  * fix KVM in the host kernel to allow guests with >40 bit address spaces
104  * (We don't want to fill all the way up to 512GB with RAM because
105  * we might want it for non-RAM purposes later. Conversely it seems
106  * reasonable to assume that anybody configuring a VM with a quarter
107  * of a terabyte of RAM will be doing it on a host with more than a
108  * terabyte of physical address space.)
109  */
110 #define RAMLIMIT_GB 255
111 #define RAMLIMIT_BYTES (RAMLIMIT_GB * 1024ULL * 1024 * 1024)
112 
113 /* Addresses and sizes of our components.
114  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
115  * 128MB..256MB is used for miscellaneous device I/O.
116  * 256MB..1GB is reserved for possible future PCI support (ie where the
117  * PCI memory window will go if we add a PCI host controller).
118  * 1GB and up is RAM (which may happily spill over into the
119  * high memory region beyond 4GB).
120  * This represents a compromise between how much RAM can be given to
121  * a 32 bit VM and leaving space for expansion and in particular for PCI.
122  * Note that devices should generally be placed at multiples of 0x10000,
123  * to accommodate guests using 64K pages.
124  */
125 static const MemMapEntry a15memmap[] = {
126     /* Space up to 0x8000000 is reserved for a boot ROM */
127     [VIRT_FLASH] =              {          0, 0x08000000 },
128     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
129     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
130     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
131     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
132     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
133     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
134     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
135     /* This redistributor space allows up to 2*64kB*123 CPUs */
136     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
137     [VIRT_UART] =               { 0x09000000, 0x00001000 },
138     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
139     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
140     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
141     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
142     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
143     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
144     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
145     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
146     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
147     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
148     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
149     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
150     [VIRT_MEM] =                { 0x40000000, RAMLIMIT_BYTES },
151     /* Second PCIe window, 512GB wide at the 512GB boundary */
152     [VIRT_PCIE_MMIO_HIGH] =   { 0x8000000000ULL, 0x8000000000ULL },
153 };
154 
155 static const int a15irqmap[] = {
156     [VIRT_UART] = 1,
157     [VIRT_RTC] = 2,
158     [VIRT_PCIE] = 3, /* ... to 6 */
159     [VIRT_GPIO] = 7,
160     [VIRT_SECURE_UART] = 8,
161     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
162     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
163     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
164     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
165 };
166 
167 static const char *valid_cpus[] = {
168     ARM_CPU_TYPE_NAME("cortex-a15"),
169     ARM_CPU_TYPE_NAME("cortex-a53"),
170     ARM_CPU_TYPE_NAME("cortex-a57"),
171     ARM_CPU_TYPE_NAME("host"),
172     ARM_CPU_TYPE_NAME("max"),
173 };
174 
175 static bool cpu_type_valid(const char *cpu)
176 {
177     int i;
178 
179     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
180         if (strcmp(cpu, valid_cpus[i]) == 0) {
181             return true;
182         }
183     }
184     return false;
185 }
186 
187 static void create_fdt(VirtMachineState *vms)
188 {
189     void *fdt = create_device_tree(&vms->fdt_size);
190 
191     if (!fdt) {
192         error_report("create_device_tree() failed");
193         exit(1);
194     }
195 
196     vms->fdt = fdt;
197 
198     /* Header */
199     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
200     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
201     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
202 
203     /*
204      * /chosen and /memory nodes must exist for load_dtb
205      * to fill in necessary properties later
206      */
207     qemu_fdt_add_subnode(fdt, "/chosen");
208     qemu_fdt_add_subnode(fdt, "/memory");
209     qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory");
210 
211     /* Clock node, for the benefit of the UART. The kernel device tree
212      * binding documentation claims the PL011 node clock properties are
213      * optional but in practice if you omit them the kernel refuses to
214      * probe for the device.
215      */
216     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
217     qemu_fdt_add_subnode(fdt, "/apb-pclk");
218     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
219     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
220     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
221     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
222                                 "clk24mhz");
223     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
224 
225     if (have_numa_distance) {
226         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
227         uint32_t *matrix = g_malloc0(size);
228         int idx, i, j;
229 
230         for (i = 0; i < nb_numa_nodes; i++) {
231             for (j = 0; j < nb_numa_nodes; j++) {
232                 idx = (i * nb_numa_nodes + j) * 3;
233                 matrix[idx + 0] = cpu_to_be32(i);
234                 matrix[idx + 1] = cpu_to_be32(j);
235                 matrix[idx + 2] = cpu_to_be32(numa_info[i].distance[j]);
236             }
237         }
238 
239         qemu_fdt_add_subnode(fdt, "/distance-map");
240         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
241                                 "numa-distance-map-v1");
242         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
243                          matrix, size);
244         g_free(matrix);
245     }
246 }
247 
248 static void fdt_add_timer_nodes(const VirtMachineState *vms)
249 {
250     /* On real hardware these interrupts are level-triggered.
251      * On KVM they were edge-triggered before host kernel version 4.4,
252      * and level-triggered afterwards.
253      * On emulated QEMU they are level-triggered.
254      *
255      * Getting the DTB info about them wrong is awkward for some
256      * guest kernels:
257      *  pre-4.8 ignore the DT and leave the interrupt configured
258      *   with whatever the GIC reset value (or the bootloader) left it at
259      *  4.8 before rc6 honour the incorrect data by programming it back
260      *   into the GIC, causing problems
261      *  4.8rc6 and later ignore the DT and always write "level triggered"
262      *   into the GIC
263      *
264      * For backwards-compatibility, virt-2.8 and earlier will continue
265      * to say these are edge-triggered, but later machines will report
266      * the correct information.
267      */
268     ARMCPU *armcpu;
269     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
270     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
271 
272     if (vmc->claim_edge_triggered_timers) {
273         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
274     }
275 
276     if (vms->gic_version == 2) {
277         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
278                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
279                              (1 << vms->smp_cpus) - 1);
280     }
281 
282     qemu_fdt_add_subnode(vms->fdt, "/timer");
283 
284     armcpu = ARM_CPU(qemu_get_cpu(0));
285     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
286         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
287         qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
288                          compat, sizeof(compat));
289     } else {
290         qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
291                                 "arm,armv7-timer");
292     }
293     qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
294     qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
295                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
296                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
297                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
298                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
299 }
300 
301 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
302 {
303     int cpu;
304     int addr_cells = 1;
305     const MachineState *ms = MACHINE(vms);
306 
307     /*
308      * From Documentation/devicetree/bindings/arm/cpus.txt
309      *  On ARM v8 64-bit systems value should be set to 2,
310      *  that corresponds to the MPIDR_EL1 register size.
311      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
312      *  in the system, #address-cells can be set to 1, since
313      *  MPIDR_EL1[63:32] bits are not used for CPUs
314      *  identification.
315      *
316      *  Here we actually don't know whether our system is 32- or 64-bit one.
317      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
318      *  at least one of them has Aff3 populated, we set #address-cells to 2.
319      */
320     for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
321         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
322 
323         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
324             addr_cells = 2;
325             break;
326         }
327     }
328 
329     qemu_fdt_add_subnode(vms->fdt, "/cpus");
330     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
331     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
332 
333     for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
334         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
335         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
336         CPUState *cs = CPU(armcpu);
337 
338         qemu_fdt_add_subnode(vms->fdt, nodename);
339         qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
340         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
341                                     armcpu->dtb_compatible);
342 
343         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
344             && vms->smp_cpus > 1) {
345             qemu_fdt_setprop_string(vms->fdt, nodename,
346                                         "enable-method", "psci");
347         }
348 
349         if (addr_cells == 2) {
350             qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
351                                  armcpu->mp_affinity);
352         } else {
353             qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
354                                   armcpu->mp_affinity);
355         }
356 
357         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
358             qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
359                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
360         }
361 
362         g_free(nodename);
363     }
364 }
365 
366 static void fdt_add_its_gic_node(VirtMachineState *vms)
367 {
368     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
369     qemu_fdt_add_subnode(vms->fdt, "/intc/its");
370     qemu_fdt_setprop_string(vms->fdt, "/intc/its", "compatible",
371                             "arm,gic-v3-its");
372     qemu_fdt_setprop(vms->fdt, "/intc/its", "msi-controller", NULL, 0);
373     qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/its", "reg",
374                                  2, vms->memmap[VIRT_GIC_ITS].base,
375                                  2, vms->memmap[VIRT_GIC_ITS].size);
376     qemu_fdt_setprop_cell(vms->fdt, "/intc/its", "phandle", vms->msi_phandle);
377 }
378 
379 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
380 {
381     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
382     qemu_fdt_add_subnode(vms->fdt, "/intc/v2m");
383     qemu_fdt_setprop_string(vms->fdt, "/intc/v2m", "compatible",
384                             "arm,gic-v2m-frame");
385     qemu_fdt_setprop(vms->fdt, "/intc/v2m", "msi-controller", NULL, 0);
386     qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/v2m", "reg",
387                                  2, vms->memmap[VIRT_GIC_V2M].base,
388                                  2, vms->memmap[VIRT_GIC_V2M].size);
389     qemu_fdt_setprop_cell(vms->fdt, "/intc/v2m", "phandle", vms->msi_phandle);
390 }
391 
392 static void fdt_add_gic_node(VirtMachineState *vms)
393 {
394     vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
395     qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
396 
397     qemu_fdt_add_subnode(vms->fdt, "/intc");
398     qemu_fdt_setprop_cell(vms->fdt, "/intc", "#interrupt-cells", 3);
399     qemu_fdt_setprop(vms->fdt, "/intc", "interrupt-controller", NULL, 0);
400     qemu_fdt_setprop_cell(vms->fdt, "/intc", "#address-cells", 0x2);
401     qemu_fdt_setprop_cell(vms->fdt, "/intc", "#size-cells", 0x2);
402     qemu_fdt_setprop(vms->fdt, "/intc", "ranges", NULL, 0);
403     if (vms->gic_version == 3) {
404         qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
405                                 "arm,gic-v3");
406         qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
407                                      2, vms->memmap[VIRT_GIC_DIST].base,
408                                      2, vms->memmap[VIRT_GIC_DIST].size,
409                                      2, vms->memmap[VIRT_GIC_REDIST].base,
410                                      2, vms->memmap[VIRT_GIC_REDIST].size);
411         if (vms->virt) {
412             qemu_fdt_setprop_cells(vms->fdt, "/intc", "interrupts",
413                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GICV3_MAINT_IRQ,
414                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
415         }
416     } else {
417         /* 'cortex-a15-gic' means 'GIC v2' */
418         qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
419                                 "arm,cortex-a15-gic");
420         qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
421                                       2, vms->memmap[VIRT_GIC_DIST].base,
422                                       2, vms->memmap[VIRT_GIC_DIST].size,
423                                       2, vms->memmap[VIRT_GIC_CPU].base,
424                                       2, vms->memmap[VIRT_GIC_CPU].size);
425     }
426 
427     qemu_fdt_setprop_cell(vms->fdt, "/intc", "phandle", vms->gic_phandle);
428 }
429 
430 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
431 {
432     CPUState *cpu;
433     ARMCPU *armcpu;
434     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
435 
436     CPU_FOREACH(cpu) {
437         armcpu = ARM_CPU(cpu);
438         if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
439             return;
440         }
441         if (kvm_enabled()) {
442             if (kvm_irqchip_in_kernel()) {
443                 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
444             }
445             kvm_arm_pmu_init(cpu);
446         }
447     }
448 
449     if (vms->gic_version == 2) {
450         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
451                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
452                              (1 << vms->smp_cpus) - 1);
453     }
454 
455     armcpu = ARM_CPU(qemu_get_cpu(0));
456     qemu_fdt_add_subnode(vms->fdt, "/pmu");
457     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
458         const char compat[] = "arm,armv8-pmuv3";
459         qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
460                          compat, sizeof(compat));
461         qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
462                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
463     }
464 }
465 
466 static void create_its(VirtMachineState *vms, DeviceState *gicdev)
467 {
468     const char *itsclass = its_class_name();
469     DeviceState *dev;
470 
471     if (!itsclass) {
472         /* Do nothing if not supported */
473         return;
474     }
475 
476     dev = qdev_create(NULL, itsclass);
477 
478     object_property_set_link(OBJECT(dev), OBJECT(gicdev), "parent-gicv3",
479                              &error_abort);
480     qdev_init_nofail(dev);
481     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
482 
483     fdt_add_its_gic_node(vms);
484 }
485 
486 static void create_v2m(VirtMachineState *vms, qemu_irq *pic)
487 {
488     int i;
489     int irq = vms->irqmap[VIRT_GIC_V2M];
490     DeviceState *dev;
491 
492     dev = qdev_create(NULL, "arm-gicv2m");
493     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
494     qdev_prop_set_uint32(dev, "base-spi", irq);
495     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
496     qdev_init_nofail(dev);
497 
498     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
499         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
500     }
501 
502     fdt_add_v2m_gic_node(vms);
503 }
504 
505 static void create_gic(VirtMachineState *vms, qemu_irq *pic)
506 {
507     /* We create a standalone GIC */
508     DeviceState *gicdev;
509     SysBusDevice *gicbusdev;
510     const char *gictype;
511     int type = vms->gic_version, i;
512 
513     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
514 
515     gicdev = qdev_create(NULL, gictype);
516     qdev_prop_set_uint32(gicdev, "revision", type);
517     qdev_prop_set_uint32(gicdev, "num-cpu", smp_cpus);
518     /* Note that the num-irq property counts both internal and external
519      * interrupts; there are always 32 of the former (mandated by GIC spec).
520      */
521     qdev_prop_set_uint32(gicdev, "num-irq", NUM_IRQS + 32);
522     if (!kvm_irqchip_in_kernel()) {
523         qdev_prop_set_bit(gicdev, "has-security-extensions", vms->secure);
524     }
525     qdev_init_nofail(gicdev);
526     gicbusdev = SYS_BUS_DEVICE(gicdev);
527     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
528     if (type == 3) {
529         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
530     } else {
531         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
532     }
533 
534     /* Wire the outputs from each CPU's generic timer and the GICv3
535      * maintenance interrupt signal to the appropriate GIC PPI inputs,
536      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
537      */
538     for (i = 0; i < smp_cpus; i++) {
539         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
540         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
541         int irq;
542         /* Mapping from the output timer irq lines from the CPU to the
543          * GIC PPI inputs we use for the virt board.
544          */
545         const int timer_irq[] = {
546             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
547             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
548             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
549             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
550         };
551 
552         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
553             qdev_connect_gpio_out(cpudev, irq,
554                                   qdev_get_gpio_in(gicdev,
555                                                    ppibase + timer_irq[irq]));
556         }
557 
558         qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt", 0,
559                                     qdev_get_gpio_in(gicdev, ppibase
560                                                      + ARCH_GICV3_MAINT_IRQ));
561         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
562                                     qdev_get_gpio_in(gicdev, ppibase
563                                                      + VIRTUAL_PMU_IRQ));
564 
565         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
566         sysbus_connect_irq(gicbusdev, i + smp_cpus,
567                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
568         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
569                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
570         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
571                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
572     }
573 
574     for (i = 0; i < NUM_IRQS; i++) {
575         pic[i] = qdev_get_gpio_in(gicdev, i);
576     }
577 
578     fdt_add_gic_node(vms);
579 
580     if (type == 3 && vms->its) {
581         create_its(vms, gicdev);
582     } else if (type == 2) {
583         create_v2m(vms, pic);
584     }
585 }
586 
587 static void create_uart(const VirtMachineState *vms, qemu_irq *pic, int uart,
588                         MemoryRegion *mem, Chardev *chr)
589 {
590     char *nodename;
591     hwaddr base = vms->memmap[uart].base;
592     hwaddr size = vms->memmap[uart].size;
593     int irq = vms->irqmap[uart];
594     const char compat[] = "arm,pl011\0arm,primecell";
595     const char clocknames[] = "uartclk\0apb_pclk";
596     DeviceState *dev = qdev_create(NULL, "pl011");
597     SysBusDevice *s = SYS_BUS_DEVICE(dev);
598 
599     qdev_prop_set_chr(dev, "chardev", chr);
600     qdev_init_nofail(dev);
601     memory_region_add_subregion(mem, base,
602                                 sysbus_mmio_get_region(s, 0));
603     sysbus_connect_irq(s, 0, pic[irq]);
604 
605     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
606     qemu_fdt_add_subnode(vms->fdt, nodename);
607     /* Note that we can't use setprop_string because of the embedded NUL */
608     qemu_fdt_setprop(vms->fdt, nodename, "compatible",
609                          compat, sizeof(compat));
610     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
611                                      2, base, 2, size);
612     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
613                                GIC_FDT_IRQ_TYPE_SPI, irq,
614                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
615     qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
616                                vms->clock_phandle, vms->clock_phandle);
617     qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
618                          clocknames, sizeof(clocknames));
619 
620     if (uart == VIRT_UART) {
621         qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
622     } else {
623         /* Mark as not usable by the normal world */
624         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
625         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
626     }
627 
628     g_free(nodename);
629 }
630 
631 static void create_rtc(const VirtMachineState *vms, qemu_irq *pic)
632 {
633     char *nodename;
634     hwaddr base = vms->memmap[VIRT_RTC].base;
635     hwaddr size = vms->memmap[VIRT_RTC].size;
636     int irq = vms->irqmap[VIRT_RTC];
637     const char compat[] = "arm,pl031\0arm,primecell";
638 
639     sysbus_create_simple("pl031", base, pic[irq]);
640 
641     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
642     qemu_fdt_add_subnode(vms->fdt, nodename);
643     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
644     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
645                                  2, base, 2, size);
646     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
647                            GIC_FDT_IRQ_TYPE_SPI, irq,
648                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
649     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
650     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
651     g_free(nodename);
652 }
653 
654 static DeviceState *gpio_key_dev;
655 static void virt_powerdown_req(Notifier *n, void *opaque)
656 {
657     /* use gpio Pin 3 for power button event */
658     qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
659 }
660 
661 static Notifier virt_system_powerdown_notifier = {
662     .notify = virt_powerdown_req
663 };
664 
665 static void create_gpio(const VirtMachineState *vms, qemu_irq *pic)
666 {
667     char *nodename;
668     DeviceState *pl061_dev;
669     hwaddr base = vms->memmap[VIRT_GPIO].base;
670     hwaddr size = vms->memmap[VIRT_GPIO].size;
671     int irq = vms->irqmap[VIRT_GPIO];
672     const char compat[] = "arm,pl061\0arm,primecell";
673 
674     pl061_dev = sysbus_create_simple("pl061", base, pic[irq]);
675 
676     uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
677     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
678     qemu_fdt_add_subnode(vms->fdt, nodename);
679     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
680                                  2, base, 2, size);
681     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
682     qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
683     qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
684     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
685                            GIC_FDT_IRQ_TYPE_SPI, irq,
686                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
687     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
688     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
689     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
690 
691     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
692                                         qdev_get_gpio_in(pl061_dev, 3));
693     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
694     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
695     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
696     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
697 
698     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
699     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
700                             "label", "GPIO Key Poweroff");
701     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
702                           KEY_POWER);
703     qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
704                            "gpios", phandle, 3, 0);
705 
706     /* connect powerdown request */
707     qemu_register_powerdown_notifier(&virt_system_powerdown_notifier);
708 
709     g_free(nodename);
710 }
711 
712 static void create_virtio_devices(const VirtMachineState *vms, qemu_irq *pic)
713 {
714     int i;
715     hwaddr size = vms->memmap[VIRT_MMIO].size;
716 
717     /* We create the transports in forwards order. Since qbus_realize()
718      * prepends (not appends) new child buses, the incrementing loop below will
719      * create a list of virtio-mmio buses with decreasing base addresses.
720      *
721      * When a -device option is processed from the command line,
722      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
723      * order. The upshot is that -device options in increasing command line
724      * order are mapped to virtio-mmio buses with decreasing base addresses.
725      *
726      * When this code was originally written, that arrangement ensured that the
727      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
728      * the first -device on the command line. (The end-to-end order is a
729      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
730      * guest kernel's name-to-address assignment strategy.)
731      *
732      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
733      * the message, if not necessarily the code, of commit 70161ff336.
734      * Therefore the loop now establishes the inverse of the original intent.
735      *
736      * Unfortunately, we can't counteract the kernel change by reversing the
737      * loop; it would break existing command lines.
738      *
739      * In any case, the kernel makes no guarantee about the stability of
740      * enumeration order of virtio devices (as demonstrated by it changing
741      * between kernel versions). For reliable and stable identification
742      * of disks users must use UUIDs or similar mechanisms.
743      */
744     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
745         int irq = vms->irqmap[VIRT_MMIO] + i;
746         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
747 
748         sysbus_create_simple("virtio-mmio", base, pic[irq]);
749     }
750 
751     /* We add dtb nodes in reverse order so that they appear in the finished
752      * device tree lowest address first.
753      *
754      * Note that this mapping is independent of the loop above. The previous
755      * loop influences virtio device to virtio transport assignment, whereas
756      * this loop controls how virtio transports are laid out in the dtb.
757      */
758     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
759         char *nodename;
760         int irq = vms->irqmap[VIRT_MMIO] + i;
761         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
762 
763         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
764         qemu_fdt_add_subnode(vms->fdt, nodename);
765         qemu_fdt_setprop_string(vms->fdt, nodename,
766                                 "compatible", "virtio,mmio");
767         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
768                                      2, base, 2, size);
769         qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
770                                GIC_FDT_IRQ_TYPE_SPI, irq,
771                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
772         qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
773         g_free(nodename);
774     }
775 }
776 
777 static void create_one_flash(const char *name, hwaddr flashbase,
778                              hwaddr flashsize, const char *file,
779                              MemoryRegion *sysmem)
780 {
781     /* Create and map a single flash device. We use the same
782      * parameters as the flash devices on the Versatile Express board.
783      */
784     DriveInfo *dinfo = drive_get_next(IF_PFLASH);
785     DeviceState *dev = qdev_create(NULL, "cfi.pflash01");
786     SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
787     const uint64_t sectorlength = 256 * 1024;
788 
789     if (dinfo) {
790         qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
791                             &error_abort);
792     }
793 
794     qdev_prop_set_uint32(dev, "num-blocks", flashsize / sectorlength);
795     qdev_prop_set_uint64(dev, "sector-length", sectorlength);
796     qdev_prop_set_uint8(dev, "width", 4);
797     qdev_prop_set_uint8(dev, "device-width", 2);
798     qdev_prop_set_bit(dev, "big-endian", false);
799     qdev_prop_set_uint16(dev, "id0", 0x89);
800     qdev_prop_set_uint16(dev, "id1", 0x18);
801     qdev_prop_set_uint16(dev, "id2", 0x00);
802     qdev_prop_set_uint16(dev, "id3", 0x00);
803     qdev_prop_set_string(dev, "name", name);
804     qdev_init_nofail(dev);
805 
806     memory_region_add_subregion(sysmem, flashbase,
807                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0));
808 
809     if (file) {
810         char *fn;
811         int image_size;
812 
813         if (drive_get(IF_PFLASH, 0, 0)) {
814             error_report("The contents of the first flash device may be "
815                          "specified with -bios or with -drive if=pflash... "
816                          "but you cannot use both options at once");
817             exit(1);
818         }
819         fn = qemu_find_file(QEMU_FILE_TYPE_BIOS, file);
820         if (!fn) {
821             error_report("Could not find ROM image '%s'", file);
822             exit(1);
823         }
824         image_size = load_image_mr(fn, sysbus_mmio_get_region(sbd, 0));
825         g_free(fn);
826         if (image_size < 0) {
827             error_report("Could not load ROM image '%s'", file);
828             exit(1);
829         }
830     }
831 }
832 
833 static void create_flash(const VirtMachineState *vms,
834                          MemoryRegion *sysmem,
835                          MemoryRegion *secure_sysmem)
836 {
837     /* Create two flash devices to fill the VIRT_FLASH space in the memmap.
838      * Any file passed via -bios goes in the first of these.
839      * sysmem is the system memory space. secure_sysmem is the secure view
840      * of the system, and the first flash device should be made visible only
841      * there. The second flash device is visible to both secure and nonsecure.
842      * If sysmem == secure_sysmem this means there is no separate Secure
843      * address space and both flash devices are generally visible.
844      */
845     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
846     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
847     char *nodename;
848 
849     create_one_flash("virt.flash0", flashbase, flashsize,
850                      bios_name, secure_sysmem);
851     create_one_flash("virt.flash1", flashbase + flashsize, flashsize,
852                      NULL, sysmem);
853 
854     if (sysmem == secure_sysmem) {
855         /* Report both flash devices as a single node in the DT */
856         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
857         qemu_fdt_add_subnode(vms->fdt, nodename);
858         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
859         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
860                                      2, flashbase, 2, flashsize,
861                                      2, flashbase + flashsize, 2, flashsize);
862         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
863         g_free(nodename);
864     } else {
865         /* Report the devices as separate nodes so we can mark one as
866          * only visible to the secure world.
867          */
868         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
869         qemu_fdt_add_subnode(vms->fdt, nodename);
870         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
871         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
872                                      2, flashbase, 2, flashsize);
873         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
874         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
875         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
876         g_free(nodename);
877 
878         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
879         qemu_fdt_add_subnode(vms->fdt, nodename);
880         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
881         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
882                                      2, flashbase + flashsize, 2, flashsize);
883         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
884         g_free(nodename);
885     }
886 }
887 
888 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
889 {
890     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
891     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
892     FWCfgState *fw_cfg;
893     char *nodename;
894 
895     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
896     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)smp_cpus);
897 
898     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
899     qemu_fdt_add_subnode(vms->fdt, nodename);
900     qemu_fdt_setprop_string(vms->fdt, nodename,
901                             "compatible", "qemu,fw-cfg-mmio");
902     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
903                                  2, base, 2, size);
904     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
905     g_free(nodename);
906     return fw_cfg;
907 }
908 
909 static void create_pcie_irq_map(const VirtMachineState *vms,
910                                 uint32_t gic_phandle,
911                                 int first_irq, const char *nodename)
912 {
913     int devfn, pin;
914     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
915     uint32_t *irq_map = full_irq_map;
916 
917     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
918         for (pin = 0; pin < 4; pin++) {
919             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
920             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
921             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
922             int i;
923 
924             uint32_t map[] = {
925                 devfn << 8, 0, 0,                           /* devfn */
926                 pin + 1,                                    /* PCI pin */
927                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
928 
929             /* Convert map to big endian */
930             for (i = 0; i < 10; i++) {
931                 irq_map[i] = cpu_to_be32(map[i]);
932             }
933             irq_map += 10;
934         }
935     }
936 
937     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
938                      full_irq_map, sizeof(full_irq_map));
939 
940     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
941                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
942                            0x7           /* PCI irq */);
943 }
944 
945 static void create_smmu(const VirtMachineState *vms, qemu_irq *pic,
946                         PCIBus *bus)
947 {
948     char *node;
949     const char compat[] = "arm,smmu-v3";
950     int irq =  vms->irqmap[VIRT_SMMU];
951     int i;
952     hwaddr base = vms->memmap[VIRT_SMMU].base;
953     hwaddr size = vms->memmap[VIRT_SMMU].size;
954     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
955     DeviceState *dev;
956 
957     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
958         return;
959     }
960 
961     dev = qdev_create(NULL, "arm-smmuv3");
962 
963     object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
964                              &error_abort);
965     qdev_init_nofail(dev);
966     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
967     for (i = 0; i < NUM_SMMU_IRQS; i++) {
968         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
969     }
970 
971     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
972     qemu_fdt_add_subnode(vms->fdt, node);
973     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
974     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
975 
976     qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
977             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
978             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
979             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
980             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
981 
982     qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
983                      sizeof(irq_names));
984 
985     qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
986     qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
987     qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
988 
989     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
990 
991     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
992     g_free(node);
993 }
994 
995 static void create_pcie(VirtMachineState *vms, qemu_irq *pic)
996 {
997     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
998     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
999     hwaddr base_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].base;
1000     hwaddr size_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].size;
1001     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1002     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1003     hwaddr base_ecam = vms->memmap[VIRT_PCIE_ECAM].base;
1004     hwaddr size_ecam = vms->memmap[VIRT_PCIE_ECAM].size;
1005     hwaddr base = base_mmio;
1006     int nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1007     int irq = vms->irqmap[VIRT_PCIE];
1008     MemoryRegion *mmio_alias;
1009     MemoryRegion *mmio_reg;
1010     MemoryRegion *ecam_alias;
1011     MemoryRegion *ecam_reg;
1012     DeviceState *dev;
1013     char *nodename;
1014     int i;
1015     PCIHostState *pci;
1016 
1017     dev = qdev_create(NULL, TYPE_GPEX_HOST);
1018     qdev_init_nofail(dev);
1019 
1020     /* Map only the first size_ecam bytes of ECAM space */
1021     ecam_alias = g_new0(MemoryRegion, 1);
1022     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1023     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1024                              ecam_reg, 0, size_ecam);
1025     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1026 
1027     /* Map the MMIO window into system address space so as to expose
1028      * the section of PCI MMIO space which starts at the same base address
1029      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1030      * the window).
1031      */
1032     mmio_alias = g_new0(MemoryRegion, 1);
1033     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1034     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1035                              mmio_reg, base_mmio, size_mmio);
1036     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1037 
1038     if (vms->highmem) {
1039         /* Map high MMIO space */
1040         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1041 
1042         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1043                                  mmio_reg, base_mmio_high, size_mmio_high);
1044         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1045                                     high_mmio_alias);
1046     }
1047 
1048     /* Map IO port space */
1049     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1050 
1051     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1052         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
1053         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1054     }
1055 
1056     pci = PCI_HOST_BRIDGE(dev);
1057     if (pci->bus) {
1058         for (i = 0; i < nb_nics; i++) {
1059             NICInfo *nd = &nd_table[i];
1060 
1061             if (!nd->model) {
1062                 nd->model = g_strdup("virtio");
1063             }
1064 
1065             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1066         }
1067     }
1068 
1069     nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1070     qemu_fdt_add_subnode(vms->fdt, nodename);
1071     qemu_fdt_setprop_string(vms->fdt, nodename,
1072                             "compatible", "pci-host-ecam-generic");
1073     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1074     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1075     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1076     qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1077     qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1078                            nr_pcie_buses - 1);
1079     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1080 
1081     if (vms->msi_phandle) {
1082         qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1083                                vms->msi_phandle);
1084     }
1085 
1086     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1087                                  2, base_ecam, 2, size_ecam);
1088 
1089     if (vms->highmem) {
1090         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1091                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1092                                      2, base_pio, 2, size_pio,
1093                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1094                                      2, base_mmio, 2, size_mmio,
1095                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1096                                      2, base_mmio_high,
1097                                      2, base_mmio_high, 2, size_mmio_high);
1098     } else {
1099         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1100                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1101                                      2, base_pio, 2, size_pio,
1102                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1103                                      2, base_mmio, 2, size_mmio);
1104     }
1105 
1106     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1107     create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1108 
1109     if (vms->iommu) {
1110         vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1111 
1112         create_smmu(vms, pic, pci->bus);
1113 
1114         qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1115                                0x0, vms->iommu_phandle, 0x0, 0x10000);
1116     }
1117 
1118     g_free(nodename);
1119 }
1120 
1121 static void create_platform_bus(VirtMachineState *vms, qemu_irq *pic)
1122 {
1123     DeviceState *dev;
1124     SysBusDevice *s;
1125     int i;
1126     MemoryRegion *sysmem = get_system_memory();
1127 
1128     dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1129     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1130     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1131     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1132     qdev_init_nofail(dev);
1133     vms->platform_bus_dev = dev;
1134 
1135     s = SYS_BUS_DEVICE(dev);
1136     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1137         int irqn = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1138         sysbus_connect_irq(s, i, pic[irqn]);
1139     }
1140 
1141     memory_region_add_subregion(sysmem,
1142                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1143                                 sysbus_mmio_get_region(s, 0));
1144 }
1145 
1146 static void create_secure_ram(VirtMachineState *vms,
1147                               MemoryRegion *secure_sysmem)
1148 {
1149     MemoryRegion *secram = g_new(MemoryRegion, 1);
1150     char *nodename;
1151     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1152     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1153 
1154     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1155                            &error_fatal);
1156     memory_region_add_subregion(secure_sysmem, base, secram);
1157 
1158     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1159     qemu_fdt_add_subnode(vms->fdt, nodename);
1160     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1161     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1162     qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1163     qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1164 
1165     g_free(nodename);
1166 }
1167 
1168 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1169 {
1170     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1171                                                  bootinfo);
1172 
1173     *fdt_size = board->fdt_size;
1174     return board->fdt;
1175 }
1176 
1177 static void virt_build_smbios(VirtMachineState *vms)
1178 {
1179     MachineClass *mc = MACHINE_GET_CLASS(vms);
1180     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1181     uint8_t *smbios_tables, *smbios_anchor;
1182     size_t smbios_tables_len, smbios_anchor_len;
1183     const char *product = "QEMU Virtual Machine";
1184 
1185     if (!vms->fw_cfg) {
1186         return;
1187     }
1188 
1189     if (kvm_enabled()) {
1190         product = "KVM Virtual Machine";
1191     }
1192 
1193     smbios_set_defaults("QEMU", product,
1194                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1195                         true, SMBIOS_ENTRY_POINT_30);
1196 
1197     smbios_get_tables(NULL, 0, &smbios_tables, &smbios_tables_len,
1198                       &smbios_anchor, &smbios_anchor_len);
1199 
1200     if (smbios_anchor) {
1201         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1202                         smbios_tables, smbios_tables_len);
1203         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1204                         smbios_anchor, smbios_anchor_len);
1205     }
1206 }
1207 
1208 static
1209 void virt_machine_done(Notifier *notifier, void *data)
1210 {
1211     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1212                                          machine_done);
1213     ARMCPU *cpu = ARM_CPU(first_cpu);
1214     struct arm_boot_info *info = &vms->bootinfo;
1215     AddressSpace *as = arm_boot_address_space(cpu, info);
1216 
1217     /*
1218      * If the user provided a dtb, we assume the dynamic sysbus nodes
1219      * already are integrated there. This corresponds to a use case where
1220      * the dynamic sysbus nodes are complex and their generation is not yet
1221      * supported. In that case the user can take charge of the guest dt
1222      * while qemu takes charge of the qom stuff.
1223      */
1224     if (info->dtb_filename == NULL) {
1225         platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1226                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1227                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1228                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1229     }
1230     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as) < 0) {
1231         exit(1);
1232     }
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     create_platform_bus(vms, pic);
1461 
1462     vms->bootinfo.ram_size = machine->ram_size;
1463     vms->bootinfo.kernel_filename = machine->kernel_filename;
1464     vms->bootinfo.kernel_cmdline = machine->kernel_cmdline;
1465     vms->bootinfo.initrd_filename = machine->initrd_filename;
1466     vms->bootinfo.nb_cpus = smp_cpus;
1467     vms->bootinfo.board_id = -1;
1468     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1469     vms->bootinfo.get_dtb = machvirt_dtb;
1470     vms->bootinfo.skip_dtb_autoload = true;
1471     vms->bootinfo.firmware_loaded = firmware_loaded;
1472     arm_load_kernel(ARM_CPU(first_cpu), &vms->bootinfo);
1473 
1474     vms->machine_done.notify = virt_machine_done;
1475     qemu_add_machine_init_done_notifier(&vms->machine_done);
1476 }
1477 
1478 static bool virt_get_secure(Object *obj, Error **errp)
1479 {
1480     VirtMachineState *vms = VIRT_MACHINE(obj);
1481 
1482     return vms->secure;
1483 }
1484 
1485 static void virt_set_secure(Object *obj, bool value, Error **errp)
1486 {
1487     VirtMachineState *vms = VIRT_MACHINE(obj);
1488 
1489     vms->secure = value;
1490 }
1491 
1492 static bool virt_get_virt(Object *obj, Error **errp)
1493 {
1494     VirtMachineState *vms = VIRT_MACHINE(obj);
1495 
1496     return vms->virt;
1497 }
1498 
1499 static void virt_set_virt(Object *obj, bool value, Error **errp)
1500 {
1501     VirtMachineState *vms = VIRT_MACHINE(obj);
1502 
1503     vms->virt = value;
1504 }
1505 
1506 static bool virt_get_highmem(Object *obj, Error **errp)
1507 {
1508     VirtMachineState *vms = VIRT_MACHINE(obj);
1509 
1510     return vms->highmem;
1511 }
1512 
1513 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1514 {
1515     VirtMachineState *vms = VIRT_MACHINE(obj);
1516 
1517     vms->highmem = value;
1518 }
1519 
1520 static bool virt_get_its(Object *obj, Error **errp)
1521 {
1522     VirtMachineState *vms = VIRT_MACHINE(obj);
1523 
1524     return vms->its;
1525 }
1526 
1527 static void virt_set_its(Object *obj, bool value, Error **errp)
1528 {
1529     VirtMachineState *vms = VIRT_MACHINE(obj);
1530 
1531     vms->its = value;
1532 }
1533 
1534 static char *virt_get_gic_version(Object *obj, Error **errp)
1535 {
1536     VirtMachineState *vms = VIRT_MACHINE(obj);
1537     const char *val = vms->gic_version == 3 ? "3" : "2";
1538 
1539     return g_strdup(val);
1540 }
1541 
1542 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1543 {
1544     VirtMachineState *vms = VIRT_MACHINE(obj);
1545 
1546     if (!strcmp(value, "3")) {
1547         vms->gic_version = 3;
1548     } else if (!strcmp(value, "2")) {
1549         vms->gic_version = 2;
1550     } else if (!strcmp(value, "host")) {
1551         vms->gic_version = 0; /* Will probe later */
1552     } else if (!strcmp(value, "max")) {
1553         vms->gic_version = -1; /* Will probe later */
1554     } else {
1555         error_setg(errp, "Invalid gic-version value");
1556         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
1557     }
1558 }
1559 
1560 static char *virt_get_iommu(Object *obj, Error **errp)
1561 {
1562     VirtMachineState *vms = VIRT_MACHINE(obj);
1563 
1564     switch (vms->iommu) {
1565     case VIRT_IOMMU_NONE:
1566         return g_strdup("none");
1567     case VIRT_IOMMU_SMMUV3:
1568         return g_strdup("smmuv3");
1569     default:
1570         g_assert_not_reached();
1571     }
1572 }
1573 
1574 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
1575 {
1576     VirtMachineState *vms = VIRT_MACHINE(obj);
1577 
1578     if (!strcmp(value, "smmuv3")) {
1579         vms->iommu = VIRT_IOMMU_SMMUV3;
1580     } else if (!strcmp(value, "none")) {
1581         vms->iommu = VIRT_IOMMU_NONE;
1582     } else {
1583         error_setg(errp, "Invalid iommu value");
1584         error_append_hint(errp, "Valid values are none, smmuv3.\n");
1585     }
1586 }
1587 
1588 static CpuInstanceProperties
1589 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
1590 {
1591     MachineClass *mc = MACHINE_GET_CLASS(ms);
1592     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1593 
1594     assert(cpu_index < possible_cpus->len);
1595     return possible_cpus->cpus[cpu_index].props;
1596 }
1597 
1598 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
1599 {
1600     return idx % nb_numa_nodes;
1601 }
1602 
1603 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
1604 {
1605     int n;
1606     VirtMachineState *vms = VIRT_MACHINE(ms);
1607 
1608     if (ms->possible_cpus) {
1609         assert(ms->possible_cpus->len == max_cpus);
1610         return ms->possible_cpus;
1611     }
1612 
1613     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1614                                   sizeof(CPUArchId) * max_cpus);
1615     ms->possible_cpus->len = max_cpus;
1616     for (n = 0; n < ms->possible_cpus->len; n++) {
1617         ms->possible_cpus->cpus[n].type = ms->cpu_type;
1618         ms->possible_cpus->cpus[n].arch_id =
1619             virt_cpu_mp_affinity(vms, n);
1620         ms->possible_cpus->cpus[n].props.has_thread_id = true;
1621         ms->possible_cpus->cpus[n].props.thread_id = n;
1622     }
1623     return ms->possible_cpus;
1624 }
1625 
1626 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1627                                         DeviceState *dev, Error **errp)
1628 {
1629     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1630 
1631     if (vms->platform_bus_dev) {
1632         if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
1633             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
1634                                      SYS_BUS_DEVICE(dev));
1635         }
1636     }
1637 }
1638 
1639 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
1640                                                         DeviceState *dev)
1641 {
1642     if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
1643         return HOTPLUG_HANDLER(machine);
1644     }
1645 
1646     return NULL;
1647 }
1648 
1649 static void virt_machine_class_init(ObjectClass *oc, void *data)
1650 {
1651     MachineClass *mc = MACHINE_CLASS(oc);
1652     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1653 
1654     mc->init = machvirt_init;
1655     /* Start max_cpus at the maximum QEMU supports. We'll further restrict
1656      * it later in machvirt_init, where we have more information about the
1657      * configuration of the particular instance.
1658      */
1659     mc->max_cpus = 255;
1660     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
1661     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
1662     mc->block_default_type = IF_VIRTIO;
1663     mc->no_cdrom = 1;
1664     mc->pci_allow_0_address = true;
1665     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
1666     mc->minimum_page_bits = 12;
1667     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
1668     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
1669     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
1670     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
1671     assert(!mc->get_hotplug_handler);
1672     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
1673     hc->plug = virt_machine_device_plug_cb;
1674 }
1675 
1676 static const TypeInfo virt_machine_info = {
1677     .name          = TYPE_VIRT_MACHINE,
1678     .parent        = TYPE_MACHINE,
1679     .abstract      = true,
1680     .instance_size = sizeof(VirtMachineState),
1681     .class_size    = sizeof(VirtMachineClass),
1682     .class_init    = virt_machine_class_init,
1683     .interfaces = (InterfaceInfo[]) {
1684          { TYPE_HOTPLUG_HANDLER },
1685          { }
1686     },
1687 };
1688 
1689 static void machvirt_machine_init(void)
1690 {
1691     type_register_static(&virt_machine_info);
1692 }
1693 type_init(machvirt_machine_init);
1694 
1695 #define VIRT_COMPAT_2_12 \
1696     HW_COMPAT_2_12
1697 
1698 static void virt_2_12_instance_init(Object *obj)
1699 {
1700     VirtMachineState *vms = VIRT_MACHINE(obj);
1701     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1702 
1703     /* EL3 is disabled by default on virt: this makes us consistent
1704      * between KVM and TCG for this board, and it also allows us to
1705      * boot UEFI blobs which assume no TrustZone support.
1706      */
1707     vms->secure = false;
1708     object_property_add_bool(obj, "secure", virt_get_secure,
1709                              virt_set_secure, NULL);
1710     object_property_set_description(obj, "secure",
1711                                     "Set on/off to enable/disable the ARM "
1712                                     "Security Extensions (TrustZone)",
1713                                     NULL);
1714 
1715     /* EL2 is also disabled by default, for similar reasons */
1716     vms->virt = false;
1717     object_property_add_bool(obj, "virtualization", virt_get_virt,
1718                              virt_set_virt, NULL);
1719     object_property_set_description(obj, "virtualization",
1720                                     "Set on/off to enable/disable emulating a "
1721                                     "guest CPU which implements the ARM "
1722                                     "Virtualization Extensions",
1723                                     NULL);
1724 
1725     /* High memory is enabled by default */
1726     vms->highmem = true;
1727     object_property_add_bool(obj, "highmem", virt_get_highmem,
1728                              virt_set_highmem, NULL);
1729     object_property_set_description(obj, "highmem",
1730                                     "Set on/off to enable/disable using "
1731                                     "physical address space above 32 bits",
1732                                     NULL);
1733     /* Default GIC type is v2 */
1734     vms->gic_version = 2;
1735     object_property_add_str(obj, "gic-version", virt_get_gic_version,
1736                         virt_set_gic_version, NULL);
1737     object_property_set_description(obj, "gic-version",
1738                                     "Set GIC version. "
1739                                     "Valid values are 2, 3 and host", NULL);
1740 
1741     if (vmc->no_its) {
1742         vms->its = false;
1743     } else {
1744         /* Default allows ITS instantiation */
1745         vms->its = true;
1746         object_property_add_bool(obj, "its", virt_get_its,
1747                                  virt_set_its, NULL);
1748         object_property_set_description(obj, "its",
1749                                         "Set on/off to enable/disable "
1750                                         "ITS instantiation",
1751                                         NULL);
1752     }
1753 
1754     /* Default disallows iommu instantiation */
1755     vms->iommu = VIRT_IOMMU_NONE;
1756     object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu, NULL);
1757     object_property_set_description(obj, "iommu",
1758                                     "Set the IOMMU type. "
1759                                     "Valid values are none and smmuv3",
1760                                     NULL);
1761 
1762     vms->memmap = a15memmap;
1763     vms->irqmap = a15irqmap;
1764 }
1765 
1766 static void virt_machine_2_12_options(MachineClass *mc)
1767 {
1768     SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_12);
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