xref: /openbmc/qemu/hw/arm/virt.c (revision e6e03dcf)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu-common.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "qapi/error.h"
36 #include "hw/sysbus.h"
37 #include "hw/boards.h"
38 #include "hw/arm/boot.h"
39 #include "hw/arm/primecell.h"
40 #include "hw/arm/virt.h"
41 #include "hw/block/flash.h"
42 #include "hw/vfio/vfio-calxeda-xgmac.h"
43 #include "hw/vfio/vfio-amd-xgbe.h"
44 #include "hw/display/ramfb.h"
45 #include "net/net.h"
46 #include "sysemu/device_tree.h"
47 #include "sysemu/numa.h"
48 #include "sysemu/runstate.h"
49 #include "sysemu/sysemu.h"
50 #include "sysemu/kvm.h"
51 #include "hw/loader.h"
52 #include "exec/address-spaces.h"
53 #include "qemu/bitops.h"
54 #include "qemu/error-report.h"
55 #include "qemu/module.h"
56 #include "hw/pci-host/gpex.h"
57 #include "hw/arm/sysbus-fdt.h"
58 #include "hw/platform-bus.h"
59 #include "hw/qdev-properties.h"
60 #include "hw/arm/fdt.h"
61 #include "hw/intc/arm_gic.h"
62 #include "hw/intc/arm_gicv3_common.h"
63 #include "hw/irq.h"
64 #include "kvm_arm.h"
65 #include "hw/firmware/smbios.h"
66 #include "qapi/visitor.h"
67 #include "standard-headers/linux/input.h"
68 #include "hw/arm/smmuv3.h"
69 #include "hw/acpi/acpi.h"
70 #include "target/arm/internals.h"
71 #include "hw/mem/pc-dimm.h"
72 #include "hw/mem/nvdimm.h"
73 #include "hw/acpi/generic_event_device.h"
74 
75 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
76     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
77                                                     void *data) \
78     { \
79         MachineClass *mc = MACHINE_CLASS(oc); \
80         virt_machine_##major##_##minor##_options(mc); \
81         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
82         if (latest) { \
83             mc->alias = "virt"; \
84         } \
85     } \
86     static const TypeInfo machvirt_##major##_##minor##_info = { \
87         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
88         .parent = TYPE_VIRT_MACHINE, \
89         .class_init = virt_##major##_##minor##_class_init, \
90     }; \
91     static void machvirt_machine_##major##_##minor##_init(void) \
92     { \
93         type_register_static(&machvirt_##major##_##minor##_info); \
94     } \
95     type_init(machvirt_machine_##major##_##minor##_init);
96 
97 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
98     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
99 #define DEFINE_VIRT_MACHINE(major, minor) \
100     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
101 
102 
103 /* Number of external interrupt lines to configure the GIC with */
104 #define NUM_IRQS 256
105 
106 #define PLATFORM_BUS_NUM_IRQS 64
107 
108 /* Legacy RAM limit in GB (< version 4.0) */
109 #define LEGACY_RAMLIMIT_GB 255
110 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
111 
112 /* Addresses and sizes of our components.
113  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
114  * 128MB..256MB is used for miscellaneous device I/O.
115  * 256MB..1GB is reserved for possible future PCI support (ie where the
116  * PCI memory window will go if we add a PCI host controller).
117  * 1GB and up is RAM (which may happily spill over into the
118  * high memory region beyond 4GB).
119  * This represents a compromise between how much RAM can be given to
120  * a 32 bit VM and leaving space for expansion and in particular for PCI.
121  * Note that devices should generally be placed at multiples of 0x10000,
122  * to accommodate guests using 64K pages.
123  */
124 static const MemMapEntry base_memmap[] = {
125     /* Space up to 0x8000000 is reserved for a boot ROM */
126     [VIRT_FLASH] =              {          0, 0x08000000 },
127     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
128     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
129     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
130     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
131     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
132     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
133     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
134     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
135     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
136     /* This redistributor space allows up to 2*64kB*123 CPUs */
137     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
138     [VIRT_UART] =               { 0x09000000, 0x00001000 },
139     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
140     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
141     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
142     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
143     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
144     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
145     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
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     /* Actual RAM size depends on initial RAM and device memory settings */
154     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
155 };
156 
157 /*
158  * Highmem IO Regions: This memory map is floating, located after the RAM.
159  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
160  * top of the RAM, so that its base get the same alignment as the size,
161  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
162  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
163  * Note the extended_memmap is sized so that it eventually also includes the
164  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
165  * index of base_memmap).
166  */
167 static MemMapEntry extended_memmap[] = {
168     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
169     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
170     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
171     /* Second PCIe window */
172     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
173 };
174 
175 static const int a15irqmap[] = {
176     [VIRT_UART] = 1,
177     [VIRT_RTC] = 2,
178     [VIRT_PCIE] = 3, /* ... to 6 */
179     [VIRT_GPIO] = 7,
180     [VIRT_SECURE_UART] = 8,
181     [VIRT_ACPI_GED] = 9,
182     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
183     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
184     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
185     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
186 };
187 
188 static const char *valid_cpus[] = {
189     ARM_CPU_TYPE_NAME("cortex-a7"),
190     ARM_CPU_TYPE_NAME("cortex-a15"),
191     ARM_CPU_TYPE_NAME("cortex-a53"),
192     ARM_CPU_TYPE_NAME("cortex-a57"),
193     ARM_CPU_TYPE_NAME("cortex-a72"),
194     ARM_CPU_TYPE_NAME("host"),
195     ARM_CPU_TYPE_NAME("max"),
196 };
197 
198 static bool cpu_type_valid(const char *cpu)
199 {
200     int i;
201 
202     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
203         if (strcmp(cpu, valid_cpus[i]) == 0) {
204             return true;
205         }
206     }
207     return false;
208 }
209 
210 static void create_fdt(VirtMachineState *vms)
211 {
212     MachineState *ms = MACHINE(vms);
213     int nb_numa_nodes = ms->numa_state->num_nodes;
214     void *fdt = create_device_tree(&vms->fdt_size);
215 
216     if (!fdt) {
217         error_report("create_device_tree() failed");
218         exit(1);
219     }
220 
221     vms->fdt = fdt;
222 
223     /* Header */
224     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
225     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
226     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
227 
228     /* /chosen must exist for load_dtb to fill in necessary properties later */
229     qemu_fdt_add_subnode(fdt, "/chosen");
230 
231     /* Clock node, for the benefit of the UART. The kernel device tree
232      * binding documentation claims the PL011 node clock properties are
233      * optional but in practice if you omit them the kernel refuses to
234      * probe for the device.
235      */
236     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
237     qemu_fdt_add_subnode(fdt, "/apb-pclk");
238     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
239     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
240     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
241     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
242                                 "clk24mhz");
243     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
244 
245     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
246         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
247         uint32_t *matrix = g_malloc0(size);
248         int idx, i, j;
249 
250         for (i = 0; i < nb_numa_nodes; i++) {
251             for (j = 0; j < nb_numa_nodes; j++) {
252                 idx = (i * nb_numa_nodes + j) * 3;
253                 matrix[idx + 0] = cpu_to_be32(i);
254                 matrix[idx + 1] = cpu_to_be32(j);
255                 matrix[idx + 2] =
256                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
257             }
258         }
259 
260         qemu_fdt_add_subnode(fdt, "/distance-map");
261         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
262                                 "numa-distance-map-v1");
263         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
264                          matrix, size);
265         g_free(matrix);
266     }
267 }
268 
269 static void fdt_add_timer_nodes(const VirtMachineState *vms)
270 {
271     /* On real hardware these interrupts are level-triggered.
272      * On KVM they were edge-triggered before host kernel version 4.4,
273      * and level-triggered afterwards.
274      * On emulated QEMU they are level-triggered.
275      *
276      * Getting the DTB info about them wrong is awkward for some
277      * guest kernels:
278      *  pre-4.8 ignore the DT and leave the interrupt configured
279      *   with whatever the GIC reset value (or the bootloader) left it at
280      *  4.8 before rc6 honour the incorrect data by programming it back
281      *   into the GIC, causing problems
282      *  4.8rc6 and later ignore the DT and always write "level triggered"
283      *   into the GIC
284      *
285      * For backwards-compatibility, virt-2.8 and earlier will continue
286      * to say these are edge-triggered, but later machines will report
287      * the correct information.
288      */
289     ARMCPU *armcpu;
290     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
291     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
292 
293     if (vmc->claim_edge_triggered_timers) {
294         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
295     }
296 
297     if (vms->gic_version == 2) {
298         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
299                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
300                              (1 << vms->smp_cpus) - 1);
301     }
302 
303     qemu_fdt_add_subnode(vms->fdt, "/timer");
304 
305     armcpu = ARM_CPU(qemu_get_cpu(0));
306     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
307         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
308         qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
309                          compat, sizeof(compat));
310     } else {
311         qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
312                                 "arm,armv7-timer");
313     }
314     qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
315     qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
316                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
317                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
318                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
319                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
320 }
321 
322 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
323 {
324     int cpu;
325     int addr_cells = 1;
326     const MachineState *ms = MACHINE(vms);
327 
328     /*
329      * From Documentation/devicetree/bindings/arm/cpus.txt
330      *  On ARM v8 64-bit systems value should be set to 2,
331      *  that corresponds to the MPIDR_EL1 register size.
332      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
333      *  in the system, #address-cells can be set to 1, since
334      *  MPIDR_EL1[63:32] bits are not used for CPUs
335      *  identification.
336      *
337      *  Here we actually don't know whether our system is 32- or 64-bit one.
338      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
339      *  at least one of them has Aff3 populated, we set #address-cells to 2.
340      */
341     for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
342         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
343 
344         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
345             addr_cells = 2;
346             break;
347         }
348     }
349 
350     qemu_fdt_add_subnode(vms->fdt, "/cpus");
351     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
352     qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
353 
354     for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
355         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
356         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
357         CPUState *cs = CPU(armcpu);
358 
359         qemu_fdt_add_subnode(vms->fdt, nodename);
360         qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
361         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
362                                     armcpu->dtb_compatible);
363 
364         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
365             && vms->smp_cpus > 1) {
366             qemu_fdt_setprop_string(vms->fdt, nodename,
367                                         "enable-method", "psci");
368         }
369 
370         if (addr_cells == 2) {
371             qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
372                                  armcpu->mp_affinity);
373         } else {
374             qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
375                                   armcpu->mp_affinity);
376         }
377 
378         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
379             qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
380                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
381         }
382 
383         g_free(nodename);
384     }
385 }
386 
387 static void fdt_add_its_gic_node(VirtMachineState *vms)
388 {
389     char *nodename;
390 
391     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
392     nodename = g_strdup_printf("/intc/its@%" PRIx64,
393                                vms->memmap[VIRT_GIC_ITS].base);
394     qemu_fdt_add_subnode(vms->fdt, nodename);
395     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
396                             "arm,gic-v3-its");
397     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
398     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
399                                  2, vms->memmap[VIRT_GIC_ITS].base,
400                                  2, vms->memmap[VIRT_GIC_ITS].size);
401     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
402     g_free(nodename);
403 }
404 
405 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
406 {
407     char *nodename;
408 
409     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
410                                vms->memmap[VIRT_GIC_V2M].base);
411     vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
412     qemu_fdt_add_subnode(vms->fdt, nodename);
413     qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
414                             "arm,gic-v2m-frame");
415     qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
416     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
417                                  2, vms->memmap[VIRT_GIC_V2M].base,
418                                  2, vms->memmap[VIRT_GIC_V2M].size);
419     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
420     g_free(nodename);
421 }
422 
423 static void fdt_add_gic_node(VirtMachineState *vms)
424 {
425     char *nodename;
426 
427     vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
428     qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
429 
430     nodename = g_strdup_printf("/intc@%" PRIx64,
431                                vms->memmap[VIRT_GIC_DIST].base);
432     qemu_fdt_add_subnode(vms->fdt, nodename);
433     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 3);
434     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-controller", NULL, 0);
435     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 0x2);
436     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 0x2);
437     qemu_fdt_setprop(vms->fdt, nodename, "ranges", NULL, 0);
438     if (vms->gic_version == 3) {
439         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
440 
441         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
442                                 "arm,gic-v3");
443 
444         qemu_fdt_setprop_cell(vms->fdt, nodename,
445                               "#redistributor-regions", nb_redist_regions);
446 
447         if (nb_redist_regions == 1) {
448             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
449                                          2, vms->memmap[VIRT_GIC_DIST].base,
450                                          2, vms->memmap[VIRT_GIC_DIST].size,
451                                          2, vms->memmap[VIRT_GIC_REDIST].base,
452                                          2, vms->memmap[VIRT_GIC_REDIST].size);
453         } else {
454             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
455                                  2, vms->memmap[VIRT_GIC_DIST].base,
456                                  2, vms->memmap[VIRT_GIC_DIST].size,
457                                  2, vms->memmap[VIRT_GIC_REDIST].base,
458                                  2, vms->memmap[VIRT_GIC_REDIST].size,
459                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
460                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
461         }
462 
463         if (vms->virt) {
464             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
465                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
466                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
467         }
468     } else {
469         /* 'cortex-a15-gic' means 'GIC v2' */
470         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
471                                 "arm,cortex-a15-gic");
472         if (!vms->virt) {
473             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
474                                          2, vms->memmap[VIRT_GIC_DIST].base,
475                                          2, vms->memmap[VIRT_GIC_DIST].size,
476                                          2, vms->memmap[VIRT_GIC_CPU].base,
477                                          2, vms->memmap[VIRT_GIC_CPU].size);
478         } else {
479             qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
480                                          2, vms->memmap[VIRT_GIC_DIST].base,
481                                          2, vms->memmap[VIRT_GIC_DIST].size,
482                                          2, vms->memmap[VIRT_GIC_CPU].base,
483                                          2, vms->memmap[VIRT_GIC_CPU].size,
484                                          2, vms->memmap[VIRT_GIC_HYP].base,
485                                          2, vms->memmap[VIRT_GIC_HYP].size,
486                                          2, vms->memmap[VIRT_GIC_VCPU].base,
487                                          2, vms->memmap[VIRT_GIC_VCPU].size);
488             qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
489                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
490                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
491         }
492     }
493 
494     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->gic_phandle);
495     g_free(nodename);
496 }
497 
498 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
499 {
500     CPUState *cpu;
501     ARMCPU *armcpu;
502     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
503 
504     CPU_FOREACH(cpu) {
505         armcpu = ARM_CPU(cpu);
506         if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
507             return;
508         }
509         if (kvm_enabled()) {
510             if (kvm_irqchip_in_kernel()) {
511                 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
512             }
513             kvm_arm_pmu_init(cpu);
514         }
515     }
516 
517     if (vms->gic_version == 2) {
518         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
519                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
520                              (1 << vms->smp_cpus) - 1);
521     }
522 
523     armcpu = ARM_CPU(qemu_get_cpu(0));
524     qemu_fdt_add_subnode(vms->fdt, "/pmu");
525     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
526         const char compat[] = "arm,armv8-pmuv3";
527         qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
528                          compat, sizeof(compat));
529         qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
530                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
531     }
532 }
533 
534 static inline DeviceState *create_acpi_ged(VirtMachineState *vms, qemu_irq *pic)
535 {
536     DeviceState *dev;
537     MachineState *ms = MACHINE(vms);
538     int irq = vms->irqmap[VIRT_ACPI_GED];
539     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
540 
541     if (ms->ram_slots) {
542         event |= ACPI_GED_MEM_HOTPLUG_EVT;
543     }
544 
545     dev = qdev_create(NULL, TYPE_ACPI_GED);
546     qdev_prop_set_uint32(dev, "ged-event", event);
547 
548     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
549     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
550     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, pic[irq]);
551 
552     qdev_init_nofail(dev);
553 
554     return dev;
555 }
556 
557 static void create_its(VirtMachineState *vms, DeviceState *gicdev)
558 {
559     const char *itsclass = its_class_name();
560     DeviceState *dev;
561 
562     if (!itsclass) {
563         /* Do nothing if not supported */
564         return;
565     }
566 
567     dev = qdev_create(NULL, itsclass);
568 
569     object_property_set_link(OBJECT(dev), OBJECT(gicdev), "parent-gicv3",
570                              &error_abort);
571     qdev_init_nofail(dev);
572     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
573 
574     fdt_add_its_gic_node(vms);
575 }
576 
577 static void create_v2m(VirtMachineState *vms, qemu_irq *pic)
578 {
579     int i;
580     int irq = vms->irqmap[VIRT_GIC_V2M];
581     DeviceState *dev;
582 
583     dev = qdev_create(NULL, "arm-gicv2m");
584     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
585     qdev_prop_set_uint32(dev, "base-spi", irq);
586     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
587     qdev_init_nofail(dev);
588 
589     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
590         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
591     }
592 
593     fdt_add_v2m_gic_node(vms);
594 }
595 
596 static void create_gic(VirtMachineState *vms, qemu_irq *pic)
597 {
598     MachineState *ms = MACHINE(vms);
599     /* We create a standalone GIC */
600     DeviceState *gicdev;
601     SysBusDevice *gicbusdev;
602     const char *gictype;
603     int type = vms->gic_version, i;
604     unsigned int smp_cpus = ms->smp.cpus;
605     uint32_t nb_redist_regions = 0;
606 
607     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
608 
609     gicdev = qdev_create(NULL, gictype);
610     qdev_prop_set_uint32(gicdev, "revision", type);
611     qdev_prop_set_uint32(gicdev, "num-cpu", smp_cpus);
612     /* Note that the num-irq property counts both internal and external
613      * interrupts; there are always 32 of the former (mandated by GIC spec).
614      */
615     qdev_prop_set_uint32(gicdev, "num-irq", NUM_IRQS + 32);
616     if (!kvm_irqchip_in_kernel()) {
617         qdev_prop_set_bit(gicdev, "has-security-extensions", vms->secure);
618     }
619 
620     if (type == 3) {
621         uint32_t redist0_capacity =
622                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
623         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
624 
625         nb_redist_regions = virt_gicv3_redist_region_count(vms);
626 
627         qdev_prop_set_uint32(gicdev, "len-redist-region-count",
628                              nb_redist_regions);
629         qdev_prop_set_uint32(gicdev, "redist-region-count[0]", redist0_count);
630 
631         if (nb_redist_regions == 2) {
632             uint32_t redist1_capacity =
633                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
634 
635             qdev_prop_set_uint32(gicdev, "redist-region-count[1]",
636                 MIN(smp_cpus - redist0_count, redist1_capacity));
637         }
638     } else {
639         if (!kvm_irqchip_in_kernel()) {
640             qdev_prop_set_bit(gicdev, "has-virtualization-extensions",
641                               vms->virt);
642         }
643     }
644     qdev_init_nofail(gicdev);
645     gicbusdev = SYS_BUS_DEVICE(gicdev);
646     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
647     if (type == 3) {
648         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
649         if (nb_redist_regions == 2) {
650             sysbus_mmio_map(gicbusdev, 2,
651                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
652         }
653     } else {
654         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
655         if (vms->virt) {
656             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
657             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
658         }
659     }
660 
661     /* Wire the outputs from each CPU's generic timer and the GICv3
662      * maintenance interrupt signal to the appropriate GIC PPI inputs,
663      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
664      */
665     for (i = 0; i < smp_cpus; i++) {
666         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
667         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
668         int irq;
669         /* Mapping from the output timer irq lines from the CPU to the
670          * GIC PPI inputs we use for the virt board.
671          */
672         const int timer_irq[] = {
673             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
674             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
675             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
676             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
677         };
678 
679         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
680             qdev_connect_gpio_out(cpudev, irq,
681                                   qdev_get_gpio_in(gicdev,
682                                                    ppibase + timer_irq[irq]));
683         }
684 
685         if (type == 3) {
686             qemu_irq irq = qdev_get_gpio_in(gicdev,
687                                             ppibase + ARCH_GIC_MAINT_IRQ);
688             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
689                                         0, irq);
690         } else if (vms->virt) {
691             qemu_irq irq = qdev_get_gpio_in(gicdev,
692                                             ppibase + ARCH_GIC_MAINT_IRQ);
693             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
694         }
695 
696         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
697                                     qdev_get_gpio_in(gicdev, ppibase
698                                                      + VIRTUAL_PMU_IRQ));
699 
700         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
701         sysbus_connect_irq(gicbusdev, i + smp_cpus,
702                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
703         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
704                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
705         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
706                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
707     }
708 
709     for (i = 0; i < NUM_IRQS; i++) {
710         pic[i] = qdev_get_gpio_in(gicdev, i);
711     }
712 
713     fdt_add_gic_node(vms);
714 
715     if (type == 3 && vms->its) {
716         create_its(vms, gicdev);
717     } else if (type == 2) {
718         create_v2m(vms, pic);
719     }
720 }
721 
722 static void create_uart(const VirtMachineState *vms, qemu_irq *pic, int uart,
723                         MemoryRegion *mem, Chardev *chr)
724 {
725     char *nodename;
726     hwaddr base = vms->memmap[uart].base;
727     hwaddr size = vms->memmap[uart].size;
728     int irq = vms->irqmap[uart];
729     const char compat[] = "arm,pl011\0arm,primecell";
730     const char clocknames[] = "uartclk\0apb_pclk";
731     DeviceState *dev = qdev_create(NULL, "pl011");
732     SysBusDevice *s = SYS_BUS_DEVICE(dev);
733 
734     qdev_prop_set_chr(dev, "chardev", chr);
735     qdev_init_nofail(dev);
736     memory_region_add_subregion(mem, base,
737                                 sysbus_mmio_get_region(s, 0));
738     sysbus_connect_irq(s, 0, pic[irq]);
739 
740     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
741     qemu_fdt_add_subnode(vms->fdt, nodename);
742     /* Note that we can't use setprop_string because of the embedded NUL */
743     qemu_fdt_setprop(vms->fdt, nodename, "compatible",
744                          compat, sizeof(compat));
745     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
746                                      2, base, 2, size);
747     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
748                                GIC_FDT_IRQ_TYPE_SPI, irq,
749                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
750     qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
751                                vms->clock_phandle, vms->clock_phandle);
752     qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
753                          clocknames, sizeof(clocknames));
754 
755     if (uart == VIRT_UART) {
756         qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
757     } else {
758         /* Mark as not usable by the normal world */
759         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
760         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
761 
762         qemu_fdt_add_subnode(vms->fdt, "/secure-chosen");
763         qemu_fdt_setprop_string(vms->fdt, "/secure-chosen", "stdout-path",
764                                 nodename);
765     }
766 
767     g_free(nodename);
768 }
769 
770 static void create_rtc(const VirtMachineState *vms, qemu_irq *pic)
771 {
772     char *nodename;
773     hwaddr base = vms->memmap[VIRT_RTC].base;
774     hwaddr size = vms->memmap[VIRT_RTC].size;
775     int irq = vms->irqmap[VIRT_RTC];
776     const char compat[] = "arm,pl031\0arm,primecell";
777 
778     sysbus_create_simple("pl031", base, pic[irq]);
779 
780     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
781     qemu_fdt_add_subnode(vms->fdt, nodename);
782     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
783     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
784                                  2, base, 2, size);
785     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
786                            GIC_FDT_IRQ_TYPE_SPI, irq,
787                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
788     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
789     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
790     g_free(nodename);
791 }
792 
793 static DeviceState *gpio_key_dev;
794 static void virt_powerdown_req(Notifier *n, void *opaque)
795 {
796     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
797 
798     if (s->acpi_dev) {
799         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
800     } else {
801         /* use gpio Pin 3 for power button event */
802         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
803     }
804 }
805 
806 static void create_gpio(const VirtMachineState *vms, qemu_irq *pic)
807 {
808     char *nodename;
809     DeviceState *pl061_dev;
810     hwaddr base = vms->memmap[VIRT_GPIO].base;
811     hwaddr size = vms->memmap[VIRT_GPIO].size;
812     int irq = vms->irqmap[VIRT_GPIO];
813     const char compat[] = "arm,pl061\0arm,primecell";
814 
815     pl061_dev = sysbus_create_simple("pl061", base, pic[irq]);
816 
817     uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
818     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
819     qemu_fdt_add_subnode(vms->fdt, nodename);
820     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
821                                  2, base, 2, size);
822     qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
823     qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
824     qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
825     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
826                            GIC_FDT_IRQ_TYPE_SPI, irq,
827                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
828     qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
829     qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
830     qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
831 
832     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
833                                         qdev_get_gpio_in(pl061_dev, 3));
834     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
835     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
836     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
837     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
838 
839     qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
840     qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
841                             "label", "GPIO Key Poweroff");
842     qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
843                           KEY_POWER);
844     qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
845                            "gpios", phandle, 3, 0);
846     g_free(nodename);
847 }
848 
849 static void create_virtio_devices(const VirtMachineState *vms, qemu_irq *pic)
850 {
851     int i;
852     hwaddr size = vms->memmap[VIRT_MMIO].size;
853 
854     /* We create the transports in forwards order. Since qbus_realize()
855      * prepends (not appends) new child buses, the incrementing loop below will
856      * create a list of virtio-mmio buses with decreasing base addresses.
857      *
858      * When a -device option is processed from the command line,
859      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
860      * order. The upshot is that -device options in increasing command line
861      * order are mapped to virtio-mmio buses with decreasing base addresses.
862      *
863      * When this code was originally written, that arrangement ensured that the
864      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
865      * the first -device on the command line. (The end-to-end order is a
866      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
867      * guest kernel's name-to-address assignment strategy.)
868      *
869      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
870      * the message, if not necessarily the code, of commit 70161ff336.
871      * Therefore the loop now establishes the inverse of the original intent.
872      *
873      * Unfortunately, we can't counteract the kernel change by reversing the
874      * loop; it would break existing command lines.
875      *
876      * In any case, the kernel makes no guarantee about the stability of
877      * enumeration order of virtio devices (as demonstrated by it changing
878      * between kernel versions). For reliable and stable identification
879      * of disks users must use UUIDs or similar mechanisms.
880      */
881     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
882         int irq = vms->irqmap[VIRT_MMIO] + i;
883         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
884 
885         sysbus_create_simple("virtio-mmio", base, pic[irq]);
886     }
887 
888     /* We add dtb nodes in reverse order so that they appear in the finished
889      * device tree lowest address first.
890      *
891      * Note that this mapping is independent of the loop above. The previous
892      * loop influences virtio device to virtio transport assignment, whereas
893      * this loop controls how virtio transports are laid out in the dtb.
894      */
895     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
896         char *nodename;
897         int irq = vms->irqmap[VIRT_MMIO] + i;
898         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
899 
900         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
901         qemu_fdt_add_subnode(vms->fdt, nodename);
902         qemu_fdt_setprop_string(vms->fdt, nodename,
903                                 "compatible", "virtio,mmio");
904         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
905                                      2, base, 2, size);
906         qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
907                                GIC_FDT_IRQ_TYPE_SPI, irq,
908                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
909         qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
910         g_free(nodename);
911     }
912 }
913 
914 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
915 
916 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
917                                         const char *name,
918                                         const char *alias_prop_name)
919 {
920     /*
921      * Create a single flash device.  We use the same parameters as
922      * the flash devices on the Versatile Express board.
923      */
924     DeviceState *dev = qdev_create(NULL, TYPE_PFLASH_CFI01);
925 
926     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
927     qdev_prop_set_uint8(dev, "width", 4);
928     qdev_prop_set_uint8(dev, "device-width", 2);
929     qdev_prop_set_bit(dev, "big-endian", false);
930     qdev_prop_set_uint16(dev, "id0", 0x89);
931     qdev_prop_set_uint16(dev, "id1", 0x18);
932     qdev_prop_set_uint16(dev, "id2", 0x00);
933     qdev_prop_set_uint16(dev, "id3", 0x00);
934     qdev_prop_set_string(dev, "name", name);
935     object_property_add_child(OBJECT(vms), name, OBJECT(dev),
936                               &error_abort);
937     object_property_add_alias(OBJECT(vms), alias_prop_name,
938                               OBJECT(dev), "drive", &error_abort);
939     return PFLASH_CFI01(dev);
940 }
941 
942 static void virt_flash_create(VirtMachineState *vms)
943 {
944     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
945     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
946 }
947 
948 static void virt_flash_map1(PFlashCFI01 *flash,
949                             hwaddr base, hwaddr size,
950                             MemoryRegion *sysmem)
951 {
952     DeviceState *dev = DEVICE(flash);
953 
954     assert(size % VIRT_FLASH_SECTOR_SIZE == 0);
955     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
956     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
957     qdev_init_nofail(dev);
958 
959     memory_region_add_subregion(sysmem, base,
960                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
961                                                        0));
962 }
963 
964 static void virt_flash_map(VirtMachineState *vms,
965                            MemoryRegion *sysmem,
966                            MemoryRegion *secure_sysmem)
967 {
968     /*
969      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
970      * sysmem is the system memory space. secure_sysmem is the secure view
971      * of the system, and the first flash device should be made visible only
972      * there. The second flash device is visible to both secure and nonsecure.
973      * If sysmem == secure_sysmem this means there is no separate Secure
974      * address space and both flash devices are generally visible.
975      */
976     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
977     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
978 
979     virt_flash_map1(vms->flash[0], flashbase, flashsize,
980                     secure_sysmem);
981     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
982                     sysmem);
983 }
984 
985 static void virt_flash_fdt(VirtMachineState *vms,
986                            MemoryRegion *sysmem,
987                            MemoryRegion *secure_sysmem)
988 {
989     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
990     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
991     char *nodename;
992 
993     if (sysmem == secure_sysmem) {
994         /* Report both flash devices as a single node in the DT */
995         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
996         qemu_fdt_add_subnode(vms->fdt, nodename);
997         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
998         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
999                                      2, flashbase, 2, flashsize,
1000                                      2, flashbase + flashsize, 2, flashsize);
1001         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1002         g_free(nodename);
1003     } else {
1004         /*
1005          * Report the devices as separate nodes so we can mark one as
1006          * only visible to the secure world.
1007          */
1008         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1009         qemu_fdt_add_subnode(vms->fdt, nodename);
1010         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1011         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1012                                      2, flashbase, 2, flashsize);
1013         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1014         qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1015         qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1016         g_free(nodename);
1017 
1018         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1019         qemu_fdt_add_subnode(vms->fdt, nodename);
1020         qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1021         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1022                                      2, flashbase + flashsize, 2, flashsize);
1023         qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1024         g_free(nodename);
1025     }
1026 }
1027 
1028 static bool virt_firmware_init(VirtMachineState *vms,
1029                                MemoryRegion *sysmem,
1030                                MemoryRegion *secure_sysmem)
1031 {
1032     int i;
1033     BlockBackend *pflash_blk0;
1034 
1035     /* Map legacy -drive if=pflash to machine properties */
1036     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1037         pflash_cfi01_legacy_drive(vms->flash[i],
1038                                   drive_get(IF_PFLASH, 0, i));
1039     }
1040 
1041     virt_flash_map(vms, sysmem, secure_sysmem);
1042 
1043     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1044 
1045     if (bios_name) {
1046         char *fname;
1047         MemoryRegion *mr;
1048         int image_size;
1049 
1050         if (pflash_blk0) {
1051             error_report("The contents of the first flash device may be "
1052                          "specified with -bios or with -drive if=pflash... "
1053                          "but you cannot use both options at once");
1054             exit(1);
1055         }
1056 
1057         /* Fall back to -bios */
1058 
1059         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1060         if (!fname) {
1061             error_report("Could not find ROM image '%s'", bios_name);
1062             exit(1);
1063         }
1064         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1065         image_size = load_image_mr(fname, mr);
1066         g_free(fname);
1067         if (image_size < 0) {
1068             error_report("Could not load ROM image '%s'", bios_name);
1069             exit(1);
1070         }
1071     }
1072 
1073     return pflash_blk0 || bios_name;
1074 }
1075 
1076 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1077 {
1078     MachineState *ms = MACHINE(vms);
1079     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1080     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1081     FWCfgState *fw_cfg;
1082     char *nodename;
1083 
1084     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1085     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1086 
1087     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1088     qemu_fdt_add_subnode(vms->fdt, nodename);
1089     qemu_fdt_setprop_string(vms->fdt, nodename,
1090                             "compatible", "qemu,fw-cfg-mmio");
1091     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1092                                  2, base, 2, size);
1093     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1094     g_free(nodename);
1095     return fw_cfg;
1096 }
1097 
1098 static void create_pcie_irq_map(const VirtMachineState *vms,
1099                                 uint32_t gic_phandle,
1100                                 int first_irq, const char *nodename)
1101 {
1102     int devfn, pin;
1103     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1104     uint32_t *irq_map = full_irq_map;
1105 
1106     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1107         for (pin = 0; pin < 4; pin++) {
1108             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1109             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1110             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1111             int i;
1112 
1113             uint32_t map[] = {
1114                 devfn << 8, 0, 0,                           /* devfn */
1115                 pin + 1,                                    /* PCI pin */
1116                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1117 
1118             /* Convert map to big endian */
1119             for (i = 0; i < 10; i++) {
1120                 irq_map[i] = cpu_to_be32(map[i]);
1121             }
1122             irq_map += 10;
1123         }
1124     }
1125 
1126     qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
1127                      full_irq_map, sizeof(full_irq_map));
1128 
1129     qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
1130                            0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
1131                            0x7           /* PCI irq */);
1132 }
1133 
1134 static void create_smmu(const VirtMachineState *vms, qemu_irq *pic,
1135                         PCIBus *bus)
1136 {
1137     char *node;
1138     const char compat[] = "arm,smmu-v3";
1139     int irq =  vms->irqmap[VIRT_SMMU];
1140     int i;
1141     hwaddr base = vms->memmap[VIRT_SMMU].base;
1142     hwaddr size = vms->memmap[VIRT_SMMU].size;
1143     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1144     DeviceState *dev;
1145 
1146     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1147         return;
1148     }
1149 
1150     dev = qdev_create(NULL, "arm-smmuv3");
1151 
1152     object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
1153                              &error_abort);
1154     qdev_init_nofail(dev);
1155     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1156     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1157         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
1158     }
1159 
1160     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1161     qemu_fdt_add_subnode(vms->fdt, node);
1162     qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1163     qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
1164 
1165     qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
1166             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1167             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1168             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1169             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1170 
1171     qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
1172                      sizeof(irq_names));
1173 
1174     qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
1175     qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
1176     qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
1177 
1178     qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1179 
1180     qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1181     g_free(node);
1182 }
1183 
1184 static void create_pcie(VirtMachineState *vms, qemu_irq *pic)
1185 {
1186     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1187     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1188     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1189     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1190     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1191     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1192     hwaddr base_ecam, size_ecam;
1193     hwaddr base = base_mmio;
1194     int nr_pcie_buses;
1195     int irq = vms->irqmap[VIRT_PCIE];
1196     MemoryRegion *mmio_alias;
1197     MemoryRegion *mmio_reg;
1198     MemoryRegion *ecam_alias;
1199     MemoryRegion *ecam_reg;
1200     DeviceState *dev;
1201     char *nodename;
1202     int i, ecam_id;
1203     PCIHostState *pci;
1204 
1205     dev = qdev_create(NULL, TYPE_GPEX_HOST);
1206     qdev_init_nofail(dev);
1207 
1208     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1209     base_ecam = vms->memmap[ecam_id].base;
1210     size_ecam = vms->memmap[ecam_id].size;
1211     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1212     /* Map only the first size_ecam bytes of ECAM space */
1213     ecam_alias = g_new0(MemoryRegion, 1);
1214     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1215     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1216                              ecam_reg, 0, size_ecam);
1217     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1218 
1219     /* Map the MMIO window into system address space so as to expose
1220      * the section of PCI MMIO space which starts at the same base address
1221      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1222      * the window).
1223      */
1224     mmio_alias = g_new0(MemoryRegion, 1);
1225     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1226     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1227                              mmio_reg, base_mmio, size_mmio);
1228     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1229 
1230     if (vms->highmem) {
1231         /* Map high MMIO space */
1232         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1233 
1234         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1235                                  mmio_reg, base_mmio_high, size_mmio_high);
1236         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1237                                     high_mmio_alias);
1238     }
1239 
1240     /* Map IO port space */
1241     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1242 
1243     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1244         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
1245         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1246     }
1247 
1248     pci = PCI_HOST_BRIDGE(dev);
1249     if (pci->bus) {
1250         for (i = 0; i < nb_nics; i++) {
1251             NICInfo *nd = &nd_table[i];
1252 
1253             if (!nd->model) {
1254                 nd->model = g_strdup("virtio");
1255             }
1256 
1257             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1258         }
1259     }
1260 
1261     nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1262     qemu_fdt_add_subnode(vms->fdt, nodename);
1263     qemu_fdt_setprop_string(vms->fdt, nodename,
1264                             "compatible", "pci-host-ecam-generic");
1265     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1266     qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1267     qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1268     qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
1269     qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1270                            nr_pcie_buses - 1);
1271     qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1272 
1273     if (vms->msi_phandle) {
1274         qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1275                                vms->msi_phandle);
1276     }
1277 
1278     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1279                                  2, base_ecam, 2, size_ecam);
1280 
1281     if (vms->highmem) {
1282         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1283                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1284                                      2, base_pio, 2, size_pio,
1285                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1286                                      2, base_mmio, 2, size_mmio,
1287                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1288                                      2, base_mmio_high,
1289                                      2, base_mmio_high, 2, size_mmio_high);
1290     } else {
1291         qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1292                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1293                                      2, base_pio, 2, size_pio,
1294                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1295                                      2, base_mmio, 2, size_mmio);
1296     }
1297 
1298     qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1299     create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1300 
1301     if (vms->iommu) {
1302         vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1303 
1304         create_smmu(vms, pic, pci->bus);
1305 
1306         qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1307                                0x0, vms->iommu_phandle, 0x0, 0x10000);
1308     }
1309 
1310     g_free(nodename);
1311 }
1312 
1313 static void create_platform_bus(VirtMachineState *vms, qemu_irq *pic)
1314 {
1315     DeviceState *dev;
1316     SysBusDevice *s;
1317     int i;
1318     MemoryRegion *sysmem = get_system_memory();
1319 
1320     dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1321     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1322     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1323     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1324     qdev_init_nofail(dev);
1325     vms->platform_bus_dev = dev;
1326 
1327     s = SYS_BUS_DEVICE(dev);
1328     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1329         int irqn = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1330         sysbus_connect_irq(s, i, pic[irqn]);
1331     }
1332 
1333     memory_region_add_subregion(sysmem,
1334                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1335                                 sysbus_mmio_get_region(s, 0));
1336 }
1337 
1338 static void create_secure_ram(VirtMachineState *vms,
1339                               MemoryRegion *secure_sysmem)
1340 {
1341     MemoryRegion *secram = g_new(MemoryRegion, 1);
1342     char *nodename;
1343     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1344     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1345 
1346     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1347                            &error_fatal);
1348     memory_region_add_subregion(secure_sysmem, base, secram);
1349 
1350     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1351     qemu_fdt_add_subnode(vms->fdt, nodename);
1352     qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1353     qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1354     qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1355     qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1356 
1357     g_free(nodename);
1358 }
1359 
1360 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1361 {
1362     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1363                                                  bootinfo);
1364 
1365     *fdt_size = board->fdt_size;
1366     return board->fdt;
1367 }
1368 
1369 static void virt_build_smbios(VirtMachineState *vms)
1370 {
1371     MachineClass *mc = MACHINE_GET_CLASS(vms);
1372     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1373     uint8_t *smbios_tables, *smbios_anchor;
1374     size_t smbios_tables_len, smbios_anchor_len;
1375     const char *product = "QEMU Virtual Machine";
1376 
1377     if (kvm_enabled()) {
1378         product = "KVM Virtual Machine";
1379     }
1380 
1381     smbios_set_defaults("QEMU", product,
1382                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1383                         true, SMBIOS_ENTRY_POINT_30);
1384 
1385     smbios_get_tables(MACHINE(vms), NULL, 0, &smbios_tables, &smbios_tables_len,
1386                       &smbios_anchor, &smbios_anchor_len);
1387 
1388     if (smbios_anchor) {
1389         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1390                         smbios_tables, smbios_tables_len);
1391         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1392                         smbios_anchor, smbios_anchor_len);
1393     }
1394 }
1395 
1396 static
1397 void virt_machine_done(Notifier *notifier, void *data)
1398 {
1399     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1400                                          machine_done);
1401     MachineState *ms = MACHINE(vms);
1402     ARMCPU *cpu = ARM_CPU(first_cpu);
1403     struct arm_boot_info *info = &vms->bootinfo;
1404     AddressSpace *as = arm_boot_address_space(cpu, info);
1405 
1406     /*
1407      * If the user provided a dtb, we assume the dynamic sysbus nodes
1408      * already are integrated there. This corresponds to a use case where
1409      * the dynamic sysbus nodes are complex and their generation is not yet
1410      * supported. In that case the user can take charge of the guest dt
1411      * while qemu takes charge of the qom stuff.
1412      */
1413     if (info->dtb_filename == NULL) {
1414         platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1415                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1416                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1417                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1418     }
1419     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1420         exit(1);
1421     }
1422 
1423     virt_acpi_setup(vms);
1424     virt_build_smbios(vms);
1425 }
1426 
1427 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1428 {
1429     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1430     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1431 
1432     if (!vmc->disallow_affinity_adjustment) {
1433         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1434          * GIC's target-list limitations. 32-bit KVM hosts currently
1435          * always create clusters of 4 CPUs, but that is expected to
1436          * change when they gain support for gicv3. When KVM is enabled
1437          * it will override the changes we make here, therefore our
1438          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1439          * and to improve SGI efficiency.
1440          */
1441         if (vms->gic_version == 3) {
1442             clustersz = GICV3_TARGETLIST_BITS;
1443         } else {
1444             clustersz = GIC_TARGETLIST_BITS;
1445         }
1446     }
1447     return arm_cpu_mp_affinity(idx, clustersz);
1448 }
1449 
1450 static void virt_set_memmap(VirtMachineState *vms)
1451 {
1452     MachineState *ms = MACHINE(vms);
1453     hwaddr base, device_memory_base, device_memory_size;
1454     int i;
1455 
1456     vms->memmap = extended_memmap;
1457 
1458     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1459         vms->memmap[i] = base_memmap[i];
1460     }
1461 
1462     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1463         error_report("unsupported number of memory slots: %"PRIu64,
1464                      ms->ram_slots);
1465         exit(EXIT_FAILURE);
1466     }
1467 
1468     /*
1469      * We compute the base of the high IO region depending on the
1470      * amount of initial and device memory. The device memory start/size
1471      * is aligned on 1GiB. We never put the high IO region below 256GiB
1472      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1473      * The device region size assumes 1GiB page max alignment per slot.
1474      */
1475     device_memory_base =
1476         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1477     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1478 
1479     /* Base address of the high IO region */
1480     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1481     if (base < device_memory_base) {
1482         error_report("maxmem/slots too huge");
1483         exit(EXIT_FAILURE);
1484     }
1485     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1486         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1487     }
1488 
1489     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1490         hwaddr size = extended_memmap[i].size;
1491 
1492         base = ROUND_UP(base, size);
1493         vms->memmap[i].base = base;
1494         vms->memmap[i].size = size;
1495         base += size;
1496     }
1497     vms->highest_gpa = base - 1;
1498     if (device_memory_size > 0) {
1499         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1500         ms->device_memory->base = device_memory_base;
1501         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1502                            "device-memory", device_memory_size);
1503     }
1504 }
1505 
1506 static void machvirt_init(MachineState *machine)
1507 {
1508     VirtMachineState *vms = VIRT_MACHINE(machine);
1509     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1510     MachineClass *mc = MACHINE_GET_CLASS(machine);
1511     const CPUArchIdList *possible_cpus;
1512     qemu_irq pic[NUM_IRQS];
1513     MemoryRegion *sysmem = get_system_memory();
1514     MemoryRegion *secure_sysmem = NULL;
1515     int n, virt_max_cpus;
1516     MemoryRegion *ram = g_new(MemoryRegion, 1);
1517     bool firmware_loaded;
1518     bool aarch64 = true;
1519     bool has_ged = !vmc->no_ged;
1520     unsigned int smp_cpus = machine->smp.cpus;
1521     unsigned int max_cpus = machine->smp.max_cpus;
1522 
1523     /*
1524      * In accelerated mode, the memory map is computed earlier in kvm_type()
1525      * to create a VM with the right number of IPA bits.
1526      */
1527     if (!vms->memmap) {
1528         virt_set_memmap(vms);
1529     }
1530 
1531     /* We can probe only here because during property set
1532      * KVM is not available yet
1533      */
1534     if (vms->gic_version <= 0) {
1535         /* "host" or "max" */
1536         if (!kvm_enabled()) {
1537             if (vms->gic_version == 0) {
1538                 error_report("gic-version=host requires KVM");
1539                 exit(1);
1540             } else {
1541                 /* "max": currently means 3 for TCG */
1542                 vms->gic_version = 3;
1543             }
1544         } else {
1545             vms->gic_version = kvm_arm_vgic_probe();
1546             if (!vms->gic_version) {
1547                 error_report(
1548                     "Unable to determine GIC version supported by host");
1549                 exit(1);
1550             }
1551         }
1552     }
1553 
1554     if (!cpu_type_valid(machine->cpu_type)) {
1555         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1556         exit(1);
1557     }
1558 
1559     if (vms->secure) {
1560         if (kvm_enabled()) {
1561             error_report("mach-virt: KVM does not support Security extensions");
1562             exit(1);
1563         }
1564 
1565         /*
1566          * The Secure view of the world is the same as the NonSecure,
1567          * but with a few extra devices. Create it as a container region
1568          * containing the system memory at low priority; any secure-only
1569          * devices go in at higher priority and take precedence.
1570          */
1571         secure_sysmem = g_new(MemoryRegion, 1);
1572         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1573                            UINT64_MAX);
1574         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1575     }
1576 
1577     firmware_loaded = virt_firmware_init(vms, sysmem,
1578                                          secure_sysmem ?: sysmem);
1579 
1580     /* If we have an EL3 boot ROM then the assumption is that it will
1581      * implement PSCI itself, so disable QEMU's internal implementation
1582      * so it doesn't get in the way. Instead of starting secondary
1583      * CPUs in PSCI powerdown state we will start them all running and
1584      * let the boot ROM sort them out.
1585      * The usual case is that we do use QEMU's PSCI implementation;
1586      * if the guest has EL2 then we will use SMC as the conduit,
1587      * and otherwise we will use HVC (for backwards compatibility and
1588      * because if we're using KVM then we must use HVC).
1589      */
1590     if (vms->secure && firmware_loaded) {
1591         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1592     } else if (vms->virt) {
1593         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1594     } else {
1595         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1596     }
1597 
1598     /* The maximum number of CPUs depends on the GIC version, or on how
1599      * many redistributors we can fit into the memory map.
1600      */
1601     if (vms->gic_version == 3) {
1602         virt_max_cpus =
1603             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1604         virt_max_cpus +=
1605             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1606     } else {
1607         virt_max_cpus = GIC_NCPU;
1608     }
1609 
1610     if (max_cpus > virt_max_cpus) {
1611         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1612                      "supported by machine 'mach-virt' (%d)",
1613                      max_cpus, virt_max_cpus);
1614         exit(1);
1615     }
1616 
1617     vms->smp_cpus = smp_cpus;
1618 
1619     if (vms->virt && kvm_enabled()) {
1620         error_report("mach-virt: KVM does not support providing "
1621                      "Virtualization extensions to the guest CPU");
1622         exit(1);
1623     }
1624 
1625     create_fdt(vms);
1626 
1627     possible_cpus = mc->possible_cpu_arch_ids(machine);
1628     for (n = 0; n < possible_cpus->len; n++) {
1629         Object *cpuobj;
1630         CPUState *cs;
1631 
1632         if (n >= smp_cpus) {
1633             break;
1634         }
1635 
1636         cpuobj = object_new(possible_cpus->cpus[n].type);
1637         object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
1638                                 "mp-affinity", NULL);
1639 
1640         cs = CPU(cpuobj);
1641         cs->cpu_index = n;
1642 
1643         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1644                           &error_fatal);
1645 
1646         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1647 
1648         if (!vms->secure) {
1649             object_property_set_bool(cpuobj, false, "has_el3", NULL);
1650         }
1651 
1652         if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
1653             object_property_set_bool(cpuobj, false, "has_el2", NULL);
1654         }
1655 
1656         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1657             object_property_set_int(cpuobj, vms->psci_conduit,
1658                                     "psci-conduit", NULL);
1659 
1660             /* Secondary CPUs start in PSCI powered-down state */
1661             if (n > 0) {
1662                 object_property_set_bool(cpuobj, true,
1663                                          "start-powered-off", NULL);
1664             }
1665         }
1666 
1667         if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1668             object_property_set_bool(cpuobj, false, "pmu", NULL);
1669         }
1670 
1671         if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1672             object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
1673                                     "reset-cbar", &error_abort);
1674         }
1675 
1676         object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1677                                  &error_abort);
1678         if (vms->secure) {
1679             object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1680                                      "secure-memory", &error_abort);
1681         }
1682 
1683         object_property_set_bool(cpuobj, true, "realized", &error_fatal);
1684         object_unref(cpuobj);
1685     }
1686     fdt_add_timer_nodes(vms);
1687     fdt_add_cpu_nodes(vms);
1688 
1689    if (!kvm_enabled()) {
1690         ARMCPU *cpu = ARM_CPU(first_cpu);
1691         bool aarch64 = object_property_get_bool(OBJECT(cpu), "aarch64", NULL);
1692 
1693         if (aarch64 && vms->highmem) {
1694             int requested_pa_size, pamax = arm_pamax(cpu);
1695 
1696             requested_pa_size = 64 - clz64(vms->highest_gpa);
1697             if (pamax < requested_pa_size) {
1698                 error_report("VCPU supports less PA bits (%d) than requested "
1699                             "by the memory map (%d)", pamax, requested_pa_size);
1700                 exit(1);
1701             }
1702         }
1703     }
1704 
1705     memory_region_allocate_system_memory(ram, NULL, "mach-virt.ram",
1706                                          machine->ram_size);
1707     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base, ram);
1708     if (machine->device_memory) {
1709         memory_region_add_subregion(sysmem, machine->device_memory->base,
1710                                     &machine->device_memory->mr);
1711     }
1712 
1713     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
1714 
1715     create_gic(vms, pic);
1716 
1717     fdt_add_pmu_nodes(vms);
1718 
1719     create_uart(vms, pic, VIRT_UART, sysmem, serial_hd(0));
1720 
1721     if (vms->secure) {
1722         create_secure_ram(vms, secure_sysmem);
1723         create_uart(vms, pic, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
1724     }
1725 
1726     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
1727 
1728     create_rtc(vms, pic);
1729 
1730     create_pcie(vms, pic);
1731 
1732     if (has_ged && aarch64 && firmware_loaded && acpi_enabled) {
1733         vms->acpi_dev = create_acpi_ged(vms, pic);
1734     } else {
1735         create_gpio(vms, pic);
1736     }
1737 
1738      /* connect powerdown request */
1739      vms->powerdown_notifier.notify = virt_powerdown_req;
1740      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
1741 
1742     /* Create mmio transports, so the user can create virtio backends
1743      * (which will be automatically plugged in to the transports). If
1744      * no backend is created the transport will just sit harmlessly idle.
1745      */
1746     create_virtio_devices(vms, pic);
1747 
1748     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1749     rom_set_fw(vms->fw_cfg);
1750 
1751     create_platform_bus(vms, pic);
1752 
1753     vms->bootinfo.ram_size = machine->ram_size;
1754     vms->bootinfo.nb_cpus = smp_cpus;
1755     vms->bootinfo.board_id = -1;
1756     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1757     vms->bootinfo.get_dtb = machvirt_dtb;
1758     vms->bootinfo.skip_dtb_autoload = true;
1759     vms->bootinfo.firmware_loaded = firmware_loaded;
1760     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
1761 
1762     vms->machine_done.notify = virt_machine_done;
1763     qemu_add_machine_init_done_notifier(&vms->machine_done);
1764 }
1765 
1766 static bool virt_get_secure(Object *obj, Error **errp)
1767 {
1768     VirtMachineState *vms = VIRT_MACHINE(obj);
1769 
1770     return vms->secure;
1771 }
1772 
1773 static void virt_set_secure(Object *obj, bool value, Error **errp)
1774 {
1775     VirtMachineState *vms = VIRT_MACHINE(obj);
1776 
1777     vms->secure = value;
1778 }
1779 
1780 static bool virt_get_virt(Object *obj, Error **errp)
1781 {
1782     VirtMachineState *vms = VIRT_MACHINE(obj);
1783 
1784     return vms->virt;
1785 }
1786 
1787 static void virt_set_virt(Object *obj, bool value, Error **errp)
1788 {
1789     VirtMachineState *vms = VIRT_MACHINE(obj);
1790 
1791     vms->virt = value;
1792 }
1793 
1794 static bool virt_get_highmem(Object *obj, Error **errp)
1795 {
1796     VirtMachineState *vms = VIRT_MACHINE(obj);
1797 
1798     return vms->highmem;
1799 }
1800 
1801 static void virt_set_highmem(Object *obj, bool value, Error **errp)
1802 {
1803     VirtMachineState *vms = VIRT_MACHINE(obj);
1804 
1805     vms->highmem = value;
1806 }
1807 
1808 static bool virt_get_its(Object *obj, Error **errp)
1809 {
1810     VirtMachineState *vms = VIRT_MACHINE(obj);
1811 
1812     return vms->its;
1813 }
1814 
1815 static void virt_set_its(Object *obj, bool value, Error **errp)
1816 {
1817     VirtMachineState *vms = VIRT_MACHINE(obj);
1818 
1819     vms->its = value;
1820 }
1821 
1822 static char *virt_get_gic_version(Object *obj, Error **errp)
1823 {
1824     VirtMachineState *vms = VIRT_MACHINE(obj);
1825     const char *val = vms->gic_version == 3 ? "3" : "2";
1826 
1827     return g_strdup(val);
1828 }
1829 
1830 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1831 {
1832     VirtMachineState *vms = VIRT_MACHINE(obj);
1833 
1834     if (!strcmp(value, "3")) {
1835         vms->gic_version = 3;
1836     } else if (!strcmp(value, "2")) {
1837         vms->gic_version = 2;
1838     } else if (!strcmp(value, "host")) {
1839         vms->gic_version = 0; /* Will probe later */
1840     } else if (!strcmp(value, "max")) {
1841         vms->gic_version = -1; /* Will probe later */
1842     } else {
1843         error_setg(errp, "Invalid gic-version value");
1844         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
1845     }
1846 }
1847 
1848 static char *virt_get_iommu(Object *obj, Error **errp)
1849 {
1850     VirtMachineState *vms = VIRT_MACHINE(obj);
1851 
1852     switch (vms->iommu) {
1853     case VIRT_IOMMU_NONE:
1854         return g_strdup("none");
1855     case VIRT_IOMMU_SMMUV3:
1856         return g_strdup("smmuv3");
1857     default:
1858         g_assert_not_reached();
1859     }
1860 }
1861 
1862 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
1863 {
1864     VirtMachineState *vms = VIRT_MACHINE(obj);
1865 
1866     if (!strcmp(value, "smmuv3")) {
1867         vms->iommu = VIRT_IOMMU_SMMUV3;
1868     } else if (!strcmp(value, "none")) {
1869         vms->iommu = VIRT_IOMMU_NONE;
1870     } else {
1871         error_setg(errp, "Invalid iommu value");
1872         error_append_hint(errp, "Valid values are none, smmuv3.\n");
1873     }
1874 }
1875 
1876 static CpuInstanceProperties
1877 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
1878 {
1879     MachineClass *mc = MACHINE_GET_CLASS(ms);
1880     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1881 
1882     assert(cpu_index < possible_cpus->len);
1883     return possible_cpus->cpus[cpu_index].props;
1884 }
1885 
1886 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
1887 {
1888     return idx % ms->numa_state->num_nodes;
1889 }
1890 
1891 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
1892 {
1893     int n;
1894     unsigned int max_cpus = ms->smp.max_cpus;
1895     VirtMachineState *vms = VIRT_MACHINE(ms);
1896 
1897     if (ms->possible_cpus) {
1898         assert(ms->possible_cpus->len == max_cpus);
1899         return ms->possible_cpus;
1900     }
1901 
1902     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1903                                   sizeof(CPUArchId) * max_cpus);
1904     ms->possible_cpus->len = max_cpus;
1905     for (n = 0; n < ms->possible_cpus->len; n++) {
1906         ms->possible_cpus->cpus[n].type = ms->cpu_type;
1907         ms->possible_cpus->cpus[n].arch_id =
1908             virt_cpu_mp_affinity(vms, n);
1909         ms->possible_cpus->cpus[n].props.has_thread_id = true;
1910         ms->possible_cpus->cpus[n].props.thread_id = n;
1911     }
1912     return ms->possible_cpus;
1913 }
1914 
1915 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1916                                  Error **errp)
1917 {
1918     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1919     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1920 
1921     if (is_nvdimm) {
1922         error_setg(errp, "nvdimm is not yet supported");
1923         return;
1924     }
1925 
1926     if (!vms->acpi_dev) {
1927         error_setg(errp,
1928                    "memory hotplug is not enabled: missing acpi-ged device");
1929         return;
1930     }
1931 
1932     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
1933 }
1934 
1935 static void virt_memory_plug(HotplugHandler *hotplug_dev,
1936                              DeviceState *dev, Error **errp)
1937 {
1938     HotplugHandlerClass *hhc;
1939     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1940     Error *local_err = NULL;
1941 
1942     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms), &local_err);
1943     if (local_err) {
1944         goto out;
1945     }
1946 
1947     hhc = HOTPLUG_HANDLER_GET_CLASS(vms->acpi_dev);
1948     hhc->plug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &error_abort);
1949 out:
1950     error_propagate(errp, local_err);
1951 }
1952 
1953 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1954                                             DeviceState *dev, Error **errp)
1955 {
1956     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1957         virt_memory_pre_plug(hotplug_dev, dev, errp);
1958     }
1959 }
1960 
1961 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1962                                         DeviceState *dev, Error **errp)
1963 {
1964     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
1965 
1966     if (vms->platform_bus_dev) {
1967         if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
1968             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
1969                                      SYS_BUS_DEVICE(dev));
1970         }
1971     }
1972     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1973         virt_memory_plug(hotplug_dev, dev, errp);
1974     }
1975 }
1976 
1977 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1978                                           DeviceState *dev, Error **errp)
1979 {
1980     error_setg(errp, "device unplug request for unsupported device"
1981                " type: %s", object_get_typename(OBJECT(dev)));
1982 }
1983 
1984 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
1985                                                         DeviceState *dev)
1986 {
1987     if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE) ||
1988        (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
1989         return HOTPLUG_HANDLER(machine);
1990     }
1991 
1992     return NULL;
1993 }
1994 
1995 /*
1996  * for arm64 kvm_type [7-0] encodes the requested number of bits
1997  * in the IPA address space
1998  */
1999 static int virt_kvm_type(MachineState *ms, const char *type_str)
2000 {
2001     VirtMachineState *vms = VIRT_MACHINE(ms);
2002     int max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms);
2003     int requested_pa_size;
2004 
2005     /* we freeze the memory map to compute the highest gpa */
2006     virt_set_memmap(vms);
2007 
2008     requested_pa_size = 64 - clz64(vms->highest_gpa);
2009 
2010     if (requested_pa_size > max_vm_pa_size) {
2011         error_report("-m and ,maxmem option values "
2012                      "require an IPA range (%d bits) larger than "
2013                      "the one supported by the host (%d bits)",
2014                      requested_pa_size, max_vm_pa_size);
2015        exit(1);
2016     }
2017     /*
2018      * By default we return 0 which corresponds to an implicit legacy
2019      * 40b IPA setting. Otherwise we return the actual requested PA
2020      * logsize
2021      */
2022     return requested_pa_size > 40 ? requested_pa_size : 0;
2023 }
2024 
2025 static void virt_machine_class_init(ObjectClass *oc, void *data)
2026 {
2027     MachineClass *mc = MACHINE_CLASS(oc);
2028     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2029 
2030     mc->init = machvirt_init;
2031     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2032      * The value may be reduced later when we have more information about the
2033      * configuration of the particular instance.
2034      */
2035     mc->max_cpus = 512;
2036     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2037     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2038     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2039     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2040     mc->block_default_type = IF_VIRTIO;
2041     mc->no_cdrom = 1;
2042     mc->pci_allow_0_address = true;
2043     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2044     mc->minimum_page_bits = 12;
2045     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2046     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2047     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2048     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2049     mc->kvm_type = virt_kvm_type;
2050     assert(!mc->get_hotplug_handler);
2051     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2052     hc->pre_plug = virt_machine_device_pre_plug_cb;
2053     hc->plug = virt_machine_device_plug_cb;
2054     hc->unplug_request = virt_machine_device_unplug_request_cb;
2055     mc->numa_mem_supported = true;
2056     mc->auto_enable_numa_with_memhp = true;
2057 }
2058 
2059 static void virt_instance_init(Object *obj)
2060 {
2061     VirtMachineState *vms = VIRT_MACHINE(obj);
2062     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2063 
2064     /* EL3 is disabled by default on virt: this makes us consistent
2065      * between KVM and TCG for this board, and it also allows us to
2066      * boot UEFI blobs which assume no TrustZone support.
2067      */
2068     vms->secure = false;
2069     object_property_add_bool(obj, "secure", virt_get_secure,
2070                              virt_set_secure, NULL);
2071     object_property_set_description(obj, "secure",
2072                                     "Set on/off to enable/disable the ARM "
2073                                     "Security Extensions (TrustZone)",
2074                                     NULL);
2075 
2076     /* EL2 is also disabled by default, for similar reasons */
2077     vms->virt = false;
2078     object_property_add_bool(obj, "virtualization", virt_get_virt,
2079                              virt_set_virt, NULL);
2080     object_property_set_description(obj, "virtualization",
2081                                     "Set on/off to enable/disable emulating a "
2082                                     "guest CPU which implements the ARM "
2083                                     "Virtualization Extensions",
2084                                     NULL);
2085 
2086     /* High memory is enabled by default */
2087     vms->highmem = true;
2088     object_property_add_bool(obj, "highmem", virt_get_highmem,
2089                              virt_set_highmem, NULL);
2090     object_property_set_description(obj, "highmem",
2091                                     "Set on/off to enable/disable using "
2092                                     "physical address space above 32 bits",
2093                                     NULL);
2094     /* Default GIC type is v2 */
2095     vms->gic_version = 2;
2096     object_property_add_str(obj, "gic-version", virt_get_gic_version,
2097                         virt_set_gic_version, NULL);
2098     object_property_set_description(obj, "gic-version",
2099                                     "Set GIC version. "
2100                                     "Valid values are 2, 3 and host", NULL);
2101 
2102     vms->highmem_ecam = !vmc->no_highmem_ecam;
2103 
2104     if (vmc->no_its) {
2105         vms->its = false;
2106     } else {
2107         /* Default allows ITS instantiation */
2108         vms->its = true;
2109         object_property_add_bool(obj, "its", virt_get_its,
2110                                  virt_set_its, NULL);
2111         object_property_set_description(obj, "its",
2112                                         "Set on/off to enable/disable "
2113                                         "ITS instantiation",
2114                                         NULL);
2115     }
2116 
2117     /* Default disallows iommu instantiation */
2118     vms->iommu = VIRT_IOMMU_NONE;
2119     object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu, NULL);
2120     object_property_set_description(obj, "iommu",
2121                                     "Set the IOMMU type. "
2122                                     "Valid values are none and smmuv3",
2123                                     NULL);
2124 
2125     vms->irqmap = a15irqmap;
2126 
2127     virt_flash_create(vms);
2128 }
2129 
2130 static const TypeInfo virt_machine_info = {
2131     .name          = TYPE_VIRT_MACHINE,
2132     .parent        = TYPE_MACHINE,
2133     .abstract      = true,
2134     .instance_size = sizeof(VirtMachineState),
2135     .class_size    = sizeof(VirtMachineClass),
2136     .class_init    = virt_machine_class_init,
2137     .instance_init = virt_instance_init,
2138     .interfaces = (InterfaceInfo[]) {
2139          { TYPE_HOTPLUG_HANDLER },
2140          { }
2141     },
2142 };
2143 
2144 static void machvirt_machine_init(void)
2145 {
2146     type_register_static(&virt_machine_info);
2147 }
2148 type_init(machvirt_machine_init);
2149 
2150 static void virt_machine_4_2_options(MachineClass *mc)
2151 {
2152 }
2153 DEFINE_VIRT_MACHINE_AS_LATEST(4, 2)
2154 
2155 static void virt_machine_4_1_options(MachineClass *mc)
2156 {
2157     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2158 
2159     virt_machine_4_2_options(mc);
2160     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2161     vmc->no_ged = true;
2162     mc->auto_enable_numa_with_memhp = false;
2163 }
2164 DEFINE_VIRT_MACHINE(4, 1)
2165 
2166 static void virt_machine_4_0_options(MachineClass *mc)
2167 {
2168     virt_machine_4_1_options(mc);
2169     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2170 }
2171 DEFINE_VIRT_MACHINE(4, 0)
2172 
2173 static void virt_machine_3_1_options(MachineClass *mc)
2174 {
2175     virt_machine_4_0_options(mc);
2176     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2177 }
2178 DEFINE_VIRT_MACHINE(3, 1)
2179 
2180 static void virt_machine_3_0_options(MachineClass *mc)
2181 {
2182     virt_machine_3_1_options(mc);
2183     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2184 }
2185 DEFINE_VIRT_MACHINE(3, 0)
2186 
2187 static void virt_machine_2_12_options(MachineClass *mc)
2188 {
2189     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2190 
2191     virt_machine_3_0_options(mc);
2192     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2193     vmc->no_highmem_ecam = true;
2194     mc->max_cpus = 255;
2195 }
2196 DEFINE_VIRT_MACHINE(2, 12)
2197 
2198 static void virt_machine_2_11_options(MachineClass *mc)
2199 {
2200     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2201 
2202     virt_machine_2_12_options(mc);
2203     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2204     vmc->smbios_old_sys_ver = true;
2205 }
2206 DEFINE_VIRT_MACHINE(2, 11)
2207 
2208 static void virt_machine_2_10_options(MachineClass *mc)
2209 {
2210     virt_machine_2_11_options(mc);
2211     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2212     /* before 2.11 we never faulted accesses to bad addresses */
2213     mc->ignore_memory_transaction_failures = true;
2214 }
2215 DEFINE_VIRT_MACHINE(2, 10)
2216 
2217 static void virt_machine_2_9_options(MachineClass *mc)
2218 {
2219     virt_machine_2_10_options(mc);
2220     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2221 }
2222 DEFINE_VIRT_MACHINE(2, 9)
2223 
2224 static void virt_machine_2_8_options(MachineClass *mc)
2225 {
2226     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2227 
2228     virt_machine_2_9_options(mc);
2229     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
2230     /* For 2.8 and earlier we falsely claimed in the DT that
2231      * our timers were edge-triggered, not level-triggered.
2232      */
2233     vmc->claim_edge_triggered_timers = true;
2234 }
2235 DEFINE_VIRT_MACHINE(2, 8)
2236 
2237 static void virt_machine_2_7_options(MachineClass *mc)
2238 {
2239     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2240 
2241     virt_machine_2_8_options(mc);
2242     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2243     /* ITS was introduced with 2.8 */
2244     vmc->no_its = true;
2245     /* Stick with 1K pages for migration compatibility */
2246     mc->minimum_page_bits = 0;
2247 }
2248 DEFINE_VIRT_MACHINE(2, 7)
2249 
2250 static void virt_machine_2_6_options(MachineClass *mc)
2251 {
2252     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2253 
2254     virt_machine_2_7_options(mc);
2255     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
2256     vmc->disallow_affinity_adjustment = true;
2257     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2258     vmc->no_pmu = true;
2259 }
2260 DEFINE_VIRT_MACHINE(2, 6)
2261