]> git.proxmox.com Git - mirror_qemu.git/blame - hw/arm/virt.c
target/arm: Cache the Tagged bit for a page in MemTxAttrs
[mirror_qemu.git] / hw / arm / virt.c
CommitLineData
f5fdcd6e
PM
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
12b16722 31#include "qemu/osdep.h"
a8d25326 32#include "qemu-common.h"
350a9c9e 33#include "qemu/units.h"
e0561e60 34#include "qemu/option.h"
70e89132 35#include "monitor/qdev.h"
da34e65c 36#include "qapi/error.h"
f5fdcd6e 37#include "hw/sysbus.h"
12e9493d 38#include "hw/boards.h"
12ec8bd5 39#include "hw/arm/boot.h"
f5fdcd6e 40#include "hw/arm/primecell.h"
afe0b380 41#include "hw/arm/virt.h"
81c7db72 42#include "hw/block/flash.h"
6f2062b9
EH
43#include "hw/vfio/vfio-calxeda-xgmac.h"
44#include "hw/vfio/vfio-amd-xgbe.h"
94692dcd 45#include "hw/display/ramfb.h"
f5fdcd6e
PM
46#include "net/net.h"
47#include "sysemu/device_tree.h"
9695200a 48#include "sysemu/numa.h"
54d31236 49#include "sysemu/runstate.h"
f5fdcd6e 50#include "sysemu/sysemu.h"
c294ac32 51#include "sysemu/tpm.h"
f5fdcd6e 52#include "sysemu/kvm.h"
acf82361 53#include "hw/loader.h"
f5fdcd6e
PM
54#include "exec/address-spaces.h"
55#include "qemu/bitops.h"
56#include "qemu/error-report.h"
0b8fa32f 57#include "qemu/module.h"
4ab29b82 58#include "hw/pci-host/gpex.h"
70e89132 59#include "hw/virtio/virtio-pci.h"
5f7a5a0e
EA
60#include "hw/arm/sysbus-fdt.h"
61#include "hw/platform-bus.h"
a27bd6c7 62#include "hw/qdev-properties.h"
decf4f80 63#include "hw/arm/fdt.h"
95eb49c8
AJ
64#include "hw/intc/arm_gic.h"
65#include "hw/intc/arm_gicv3_common.h"
64552b6b 66#include "hw/irq.h"
e6fbcbc4 67#include "kvm_arm.h"
a2eb5c0c 68#include "hw/firmware/smbios.h"
b92ad394 69#include "qapi/visitor.h"
17e89077 70#include "qapi/qapi-visit-common.h"
3e6ebb64 71#include "standard-headers/linux/input.h"
584105ea 72#include "hw/arm/smmuv3.h"
957e32cf 73#include "hw/acpi/acpi.h"
2ba956cc 74#include "target/arm/internals.h"
1f283ae1
EA
75#include "hw/mem/pc-dimm.h"
76#include "hw/mem/nvdimm.h"
cff51ac9 77#include "hw/acpi/generic_event_device.h"
70e89132 78#include "hw/virtio/virtio-iommu.h"
d8f6d15f 79#include "hw/char/pl011.h"
60592cfe 80#include "qemu/guest-random.h"
f5fdcd6e 81
3356ebce 82#define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
ab093c3c
AJ
83 static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
84 void *data) \
85 { \
86 MachineClass *mc = MACHINE_CLASS(oc); \
87 virt_machine_##major##_##minor##_options(mc); \
88 mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
3356ebce
AJ
89 if (latest) { \
90 mc->alias = "virt"; \
91 } \
ab093c3c
AJ
92 } \
93 static const TypeInfo machvirt_##major##_##minor##_info = { \
94 .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
95 .parent = TYPE_VIRT_MACHINE, \
ab093c3c
AJ
96 .class_init = virt_##major##_##minor##_class_init, \
97 }; \
98 static void machvirt_machine_##major##_##minor##_init(void) \
99 { \
100 type_register_static(&machvirt_##major##_##minor##_info); \
101 } \
102 type_init(machvirt_machine_##major##_##minor##_init);
103
3356ebce
AJ
104#define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
105 DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
106#define DEFINE_VIRT_MACHINE(major, minor) \
107 DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
108
ab093c3c 109
a72d4363
AJ
110/* Number of external interrupt lines to configure the GIC with */
111#define NUM_IRQS 256
112
113#define PLATFORM_BUS_NUM_IRQS 64
114
50a17297 115/* Legacy RAM limit in GB (< version 4.0) */
957e32cf
EA
116#define LEGACY_RAMLIMIT_GB 255
117#define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
71c27684 118
f5fdcd6e
PM
119/* Addresses and sizes of our components.
120 * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
121 * 128MB..256MB is used for miscellaneous device I/O.
122 * 256MB..1GB is reserved for possible future PCI support (ie where the
123 * PCI memory window will go if we add a PCI host controller).
124 * 1GB and up is RAM (which may happily spill over into the
125 * high memory region beyond 4GB).
126 * This represents a compromise between how much RAM can be given to
127 * a 32 bit VM and leaving space for expansion and in particular for PCI.
6e411af9
PM
128 * Note that devices should generally be placed at multiples of 0x10000,
129 * to accommodate guests using 64K pages.
f5fdcd6e 130 */
350a9c9e 131static const MemMapEntry base_memmap[] = {
f5fdcd6e 132 /* Space up to 0x8000000 is reserved for a boot ROM */
94edf02c
EA
133 [VIRT_FLASH] = { 0, 0x08000000 },
134 [VIRT_CPUPERIPHS] = { 0x08000000, 0x00020000 },
f5fdcd6e 135 /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
94edf02c
EA
136 [VIRT_GIC_DIST] = { 0x08000000, 0x00010000 },
137 [VIRT_GIC_CPU] = { 0x08010000, 0x00010000 },
138 [VIRT_GIC_V2M] = { 0x08020000, 0x00001000 },
55ef3233
LM
139 [VIRT_GIC_HYP] = { 0x08030000, 0x00010000 },
140 [VIRT_GIC_VCPU] = { 0x08040000, 0x00010000 },
b92ad394
PF
141 /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
142 [VIRT_GIC_ITS] = { 0x08080000, 0x00020000 },
143 /* This redistributor space allows up to 2*64kB*123 CPUs */
144 [VIRT_GIC_REDIST] = { 0x080A0000, 0x00F60000 },
94edf02c
EA
145 [VIRT_UART] = { 0x09000000, 0x00001000 },
146 [VIRT_RTC] = { 0x09010000, 0x00001000 },
0b341a85 147 [VIRT_FW_CFG] = { 0x09020000, 0x00000018 },
b0a3721e 148 [VIRT_GPIO] = { 0x09030000, 0x00001000 },
3df708eb 149 [VIRT_SECURE_UART] = { 0x09040000, 0x00001000 },
584105ea 150 [VIRT_SMMU] = { 0x09050000, 0x00020000 },
cff51ac9
SK
151 [VIRT_PCDIMM_ACPI] = { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
152 [VIRT_ACPI_GED] = { 0x09080000, ACPI_GED_EVT_SEL_LEN },
b5a60bee 153 [VIRT_NVDIMM_ACPI] = { 0x09090000, NVDIMM_ACPI_IO_LEN},
94edf02c 154 [VIRT_MMIO] = { 0x0a000000, 0x00000200 },
f5fdcd6e 155 /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
94edf02c 156 [VIRT_PLATFORM_BUS] = { 0x0c000000, 0x02000000 },
83ec1923 157 [VIRT_SECURE_MEM] = { 0x0e000000, 0x01000000 },
94edf02c
EA
158 [VIRT_PCIE_MMIO] = { 0x10000000, 0x2eff0000 },
159 [VIRT_PCIE_PIO] = { 0x3eff0000, 0x00010000 },
160 [VIRT_PCIE_ECAM] = { 0x3f000000, 0x01000000 },
957e32cf
EA
161 /* Actual RAM size depends on initial RAM and device memory settings */
162 [VIRT_MEM] = { GiB, LEGACY_RAMLIMIT_BYTES },
350a9c9e
EA
163};
164
165/*
166 * Highmem IO Regions: This memory map is floating, located after the RAM.
167 * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
168 * top of the RAM, so that its base get the same alignment as the size,
169 * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
170 * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
171 * Note the extended_memmap is sized so that it eventually also includes the
172 * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
173 * index of base_memmap).
174 */
175static MemMapEntry extended_memmap[] = {
f90747c4 176 /* Additional 64 MB redist region (can contain up to 512 redistributors) */
350a9c9e
EA
177 [VIRT_HIGH_GIC_REDIST2] = { 0x0, 64 * MiB },
178 [VIRT_HIGH_PCIE_ECAM] = { 0x0, 256 * MiB },
179 /* Second PCIe window */
180 [VIRT_HIGH_PCIE_MMIO] = { 0x0, 512 * GiB },
f5fdcd6e
PM
181};
182
183static const int a15irqmap[] = {
184 [VIRT_UART] = 1,
6e411af9 185 [VIRT_RTC] = 2,
4ab29b82 186 [VIRT_PCIE] = 3, /* ... to 6 */
b0a3721e 187 [VIRT_GPIO] = 7,
3df708eb 188 [VIRT_SECURE_UART] = 8,
cff51ac9 189 [VIRT_ACPI_GED] = 9,
f5fdcd6e 190 [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
bd204e63 191 [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
584105ea 192 [VIRT_SMMU] = 74, /* ...to 74 + NUM_SMMU_IRQS - 1 */
5f7a5a0e 193 [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
f5fdcd6e
PM
194};
195
9ac4ef77 196static const char *valid_cpus[] = {
4414942e 197 ARM_CPU_TYPE_NAME("cortex-a7"),
ba1ba5cc
IM
198 ARM_CPU_TYPE_NAME("cortex-a15"),
199 ARM_CPU_TYPE_NAME("cortex-a53"),
200 ARM_CPU_TYPE_NAME("cortex-a57"),
2264faa5 201 ARM_CPU_TYPE_NAME("cortex-a72"),
ba1ba5cc 202 ARM_CPU_TYPE_NAME("host"),
9076ddb3 203 ARM_CPU_TYPE_NAME("max"),
f5fdcd6e
PM
204};
205
ba1ba5cc 206static bool cpu_type_valid(const char *cpu)
f5fdcd6e
PM
207{
208 int i;
209
9ac4ef77
PM
210 for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
211 if (strcmp(cpu, valid_cpus[i]) == 0) {
212 return true;
f5fdcd6e
PM
213 }
214 }
9ac4ef77 215 return false;
f5fdcd6e
PM
216}
217
60592cfe
JF
218static void create_kaslr_seed(VirtMachineState *vms, const char *node)
219{
220 Error *err = NULL;
221 uint64_t seed;
222
223 if (qemu_guest_getrandom(&seed, sizeof(seed), &err)) {
224 error_free(err);
225 return;
226 }
227 qemu_fdt_setprop_u64(vms->fdt, node, "kaslr-seed", seed);
228}
229
c8ef2bda 230static void create_fdt(VirtMachineState *vms)
f5fdcd6e 231{
aa570207
TX
232 MachineState *ms = MACHINE(vms);
233 int nb_numa_nodes = ms->numa_state->num_nodes;
c8ef2bda 234 void *fdt = create_device_tree(&vms->fdt_size);
f5fdcd6e
PM
235
236 if (!fdt) {
237 error_report("create_device_tree() failed");
238 exit(1);
239 }
240
c8ef2bda 241 vms->fdt = fdt;
f5fdcd6e
PM
242
243 /* Header */
5a4348d1
PC
244 qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
245 qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
246 qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
f5fdcd6e 247
e2eb3d29 248 /* /chosen must exist for load_dtb to fill in necessary properties later */
5a4348d1 249 qemu_fdt_add_subnode(fdt, "/chosen");
60592cfe 250 create_kaslr_seed(vms, "/chosen");
f5fdcd6e 251
ef6a5c71
JF
252 if (vms->secure) {
253 qemu_fdt_add_subnode(fdt, "/secure-chosen");
60592cfe 254 create_kaslr_seed(vms, "/secure-chosen");
ef6a5c71
JF
255 }
256
f5fdcd6e
PM
257 /* Clock node, for the benefit of the UART. The kernel device tree
258 * binding documentation claims the PL011 node clock properties are
259 * optional but in practice if you omit them the kernel refuses to
260 * probe for the device.
261 */
c8ef2bda 262 vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
5a4348d1
PC
263 qemu_fdt_add_subnode(fdt, "/apb-pclk");
264 qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
265 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
266 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
267 qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
f5fdcd6e 268 "clk24mhz");
c8ef2bda 269 qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
f5fdcd6e 270
118154b7 271 if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
c7637c04
AJ
272 int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
273 uint32_t *matrix = g_malloc0(size);
274 int idx, i, j;
275
276 for (i = 0; i < nb_numa_nodes; i++) {
277 for (j = 0; j < nb_numa_nodes; j++) {
278 idx = (i * nb_numa_nodes + j) * 3;
279 matrix[idx + 0] = cpu_to_be32(i);
280 matrix[idx + 1] = cpu_to_be32(j);
7e721e7b
TX
281 matrix[idx + 2] =
282 cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
c7637c04
AJ
283 }
284 }
285
286 qemu_fdt_add_subnode(fdt, "/distance-map");
287 qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
288 "numa-distance-map-v1");
289 qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
290 matrix, size);
291 g_free(matrix);
292 }
06955739
PS
293}
294
055a7f2b 295static void fdt_add_timer_nodes(const VirtMachineState *vms)
f5fdcd6e 296{
156bc9a5
PM
297 /* On real hardware these interrupts are level-triggered.
298 * On KVM they were edge-triggered before host kernel version 4.4,
299 * and level-triggered afterwards.
300 * On emulated QEMU they are level-triggered.
301 *
302 * Getting the DTB info about them wrong is awkward for some
303 * guest kernels:
304 * pre-4.8 ignore the DT and leave the interrupt configured
305 * with whatever the GIC reset value (or the bootloader) left it at
306 * 4.8 before rc6 honour the incorrect data by programming it back
307 * into the GIC, causing problems
308 * 4.8rc6 and later ignore the DT and always write "level triggered"
309 * into the GIC
310 *
311 * For backwards-compatibility, virt-2.8 and earlier will continue
312 * to say these are edge-triggered, but later machines will report
313 * the correct information.
f5fdcd6e 314 */
b32a9509 315 ARMCPU *armcpu;
156bc9a5
PM
316 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
317 uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
318
319 if (vmc->claim_edge_triggered_timers) {
320 irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
321 }
f5fdcd6e 322
d04460e5 323 if (vms->gic_version == VIRT_GIC_VERSION_2) {
b92ad394
PF
324 irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
325 GIC_FDT_IRQ_PPI_CPU_WIDTH,
c8ef2bda 326 (1 << vms->smp_cpus) - 1);
b92ad394 327 }
f5fdcd6e 328
c8ef2bda 329 qemu_fdt_add_subnode(vms->fdt, "/timer");
b32a9509
CF
330
331 armcpu = ARM_CPU(qemu_get_cpu(0));
332 if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
333 const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
c8ef2bda 334 qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
b32a9509
CF
335 compat, sizeof(compat));
336 } else {
c8ef2bda 337 qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
b32a9509
CF
338 "arm,armv7-timer");
339 }
c8ef2bda
PM
340 qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
341 qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
ee246400
SZ
342 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
343 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
344 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
345 GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
f5fdcd6e
PM
346}
347
c8ef2bda 348static void fdt_add_cpu_nodes(const VirtMachineState *vms)
f5fdcd6e
PM
349{
350 int cpu;
8d45c54d 351 int addr_cells = 1;
4ccf5826 352 const MachineState *ms = MACHINE(vms);
8d45c54d
PF
353
354 /*
355 * From Documentation/devicetree/bindings/arm/cpus.txt
356 * On ARM v8 64-bit systems value should be set to 2,
357 * that corresponds to the MPIDR_EL1 register size.
358 * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
359 * in the system, #address-cells can be set to 1, since
360 * MPIDR_EL1[63:32] bits are not used for CPUs
361 * identification.
362 *
363 * Here we actually don't know whether our system is 32- or 64-bit one.
364 * The simplest way to go is to examine affinity IDs of all our CPUs. If
365 * at least one of them has Aff3 populated, we set #address-cells to 2.
366 */
c8ef2bda 367 for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
8d45c54d
PF
368 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
369
370 if (armcpu->mp_affinity & ARM_AFF3_MASK) {
371 addr_cells = 2;
372 break;
373 }
374 }
f5fdcd6e 375
c8ef2bda
PM
376 qemu_fdt_add_subnode(vms->fdt, "/cpus");
377 qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
378 qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
f5fdcd6e 379
c8ef2bda 380 for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
f5fdcd6e
PM
381 char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
382 ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
4ccf5826 383 CPUState *cs = CPU(armcpu);
f5fdcd6e 384
c8ef2bda
PM
385 qemu_fdt_add_subnode(vms->fdt, nodename);
386 qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
387 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
f5fdcd6e
PM
388 armcpu->dtb_compatible);
389
2013c566
PM
390 if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
391 && vms->smp_cpus > 1) {
c8ef2bda 392 qemu_fdt_setprop_string(vms->fdt, nodename,
f5fdcd6e
PM
393 "enable-method", "psci");
394 }
395
8d45c54d 396 if (addr_cells == 2) {
c8ef2bda 397 qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
8d45c54d
PF
398 armcpu->mp_affinity);
399 } else {
c8ef2bda 400 qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
8d45c54d
PF
401 armcpu->mp_affinity);
402 }
403
4ccf5826
IM
404 if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
405 qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
406 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
9695200a
SZ
407 }
408
f5fdcd6e
PM
409 g_free(nodename);
410 }
411}
412
c8ef2bda 413static void fdt_add_its_gic_node(VirtMachineState *vms)
02f98731 414{
bb2a3348
EA
415 char *nodename;
416
c8ef2bda 417 vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
bb2a3348
EA
418 nodename = g_strdup_printf("/intc/its@%" PRIx64,
419 vms->memmap[VIRT_GIC_ITS].base);
420 qemu_fdt_add_subnode(vms->fdt, nodename);
421 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
02f98731 422 "arm,gic-v3-its");
bb2a3348
EA
423 qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
424 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
c8ef2bda
PM
425 2, vms->memmap[VIRT_GIC_ITS].base,
426 2, vms->memmap[VIRT_GIC_ITS].size);
bb2a3348
EA
427 qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
428 g_free(nodename);
02f98731
PF
429}
430
c8ef2bda 431static void fdt_add_v2m_gic_node(VirtMachineState *vms)
f5fdcd6e 432{
bb2a3348
EA
433 char *nodename;
434
435 nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
436 vms->memmap[VIRT_GIC_V2M].base);
c8ef2bda 437 vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
bb2a3348
EA
438 qemu_fdt_add_subnode(vms->fdt, nodename);
439 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
bd204e63 440 "arm,gic-v2m-frame");
bb2a3348
EA
441 qemu_fdt_setprop(vms->fdt, nodename, "msi-controller", NULL, 0);
442 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
c8ef2bda
PM
443 2, vms->memmap[VIRT_GIC_V2M].base,
444 2, vms->memmap[VIRT_GIC_V2M].size);
bb2a3348
EA
445 qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->msi_phandle);
446 g_free(nodename);
bd204e63 447}
f5fdcd6e 448
055a7f2b 449static void fdt_add_gic_node(VirtMachineState *vms)
bd204e63 450{
bb2a3348
EA
451 char *nodename;
452
c8ef2bda
PM
453 vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
454 qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
455
bb2a3348
EA
456 nodename = g_strdup_printf("/intc@%" PRIx64,
457 vms->memmap[VIRT_GIC_DIST].base);
458 qemu_fdt_add_subnode(vms->fdt, nodename);
459 qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 3);
460 qemu_fdt_setprop(vms->fdt, nodename, "interrupt-controller", NULL, 0);
461 qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 0x2);
462 qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 0x2);
463 qemu_fdt_setprop(vms->fdt, nodename, "ranges", NULL, 0);
d04460e5 464 if (vms->gic_version == VIRT_GIC_VERSION_3) {
f90747c4
EA
465 int nb_redist_regions = virt_gicv3_redist_region_count(vms);
466
bb2a3348 467 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
b92ad394 468 "arm,gic-v3");
f90747c4 469
bb2a3348 470 qemu_fdt_setprop_cell(vms->fdt, nodename,
f90747c4
EA
471 "#redistributor-regions", nb_redist_regions);
472
473 if (nb_redist_regions == 1) {
bb2a3348 474 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
f90747c4
EA
475 2, vms->memmap[VIRT_GIC_DIST].base,
476 2, vms->memmap[VIRT_GIC_DIST].size,
477 2, vms->memmap[VIRT_GIC_REDIST].base,
478 2, vms->memmap[VIRT_GIC_REDIST].size);
479 } else {
bb2a3348 480 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
bf424a12
EA
481 2, vms->memmap[VIRT_GIC_DIST].base,
482 2, vms->memmap[VIRT_GIC_DIST].size,
483 2, vms->memmap[VIRT_GIC_REDIST].base,
484 2, vms->memmap[VIRT_GIC_REDIST].size,
485 2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
486 2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
f90747c4
EA
487 }
488
f29cacfb 489 if (vms->virt) {
bb2a3348 490 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
55ef3233 491 GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
f29cacfb
PM
492 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
493 }
b92ad394
PF
494 } else {
495 /* 'cortex-a15-gic' means 'GIC v2' */
bb2a3348 496 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
b92ad394 497 "arm,cortex-a15-gic");
55ef3233
LM
498 if (!vms->virt) {
499 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
500 2, vms->memmap[VIRT_GIC_DIST].base,
501 2, vms->memmap[VIRT_GIC_DIST].size,
502 2, vms->memmap[VIRT_GIC_CPU].base,
503 2, vms->memmap[VIRT_GIC_CPU].size);
504 } else {
505 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
506 2, vms->memmap[VIRT_GIC_DIST].base,
507 2, vms->memmap[VIRT_GIC_DIST].size,
508 2, vms->memmap[VIRT_GIC_CPU].base,
509 2, vms->memmap[VIRT_GIC_CPU].size,
510 2, vms->memmap[VIRT_GIC_HYP].base,
511 2, vms->memmap[VIRT_GIC_HYP].size,
512 2, vms->memmap[VIRT_GIC_VCPU].base,
513 2, vms->memmap[VIRT_GIC_VCPU].size);
514 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
515 GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
516 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
517 }
b92ad394
PF
518 }
519
bb2a3348
EA
520 qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", vms->gic_phandle);
521 g_free(nodename);
f5fdcd6e
PM
522}
523
055a7f2b 524static void fdt_add_pmu_nodes(const VirtMachineState *vms)
01fe6b60
SZ
525{
526 CPUState *cpu;
527 ARMCPU *armcpu;
528 uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
529
530 CPU_FOREACH(cpu) {
531 armcpu = ARM_CPU(cpu);
3f07cb2a 532 if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
01fe6b60
SZ
533 return;
534 }
3f07cb2a 535 if (kvm_enabled()) {
b2bfe9f7
AJ
536 if (kvm_irqchip_in_kernel()) {
537 kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
3f07cb2a 538 }
b2bfe9f7 539 kvm_arm_pmu_init(cpu);
3f07cb2a 540 }
01fe6b60
SZ
541 }
542
d04460e5 543 if (vms->gic_version == VIRT_GIC_VERSION_2) {
01fe6b60
SZ
544 irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
545 GIC_FDT_IRQ_PPI_CPU_WIDTH,
c8ef2bda 546 (1 << vms->smp_cpus) - 1);
01fe6b60
SZ
547 }
548
549 armcpu = ARM_CPU(qemu_get_cpu(0));
c8ef2bda 550 qemu_fdt_add_subnode(vms->fdt, "/pmu");
01fe6b60
SZ
551 if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
552 const char compat[] = "arm,armv8-pmuv3";
c8ef2bda 553 qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
01fe6b60 554 compat, sizeof(compat));
c8ef2bda 555 qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
01fe6b60
SZ
556 GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
557 }
558}
559
b8b69f4c 560static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
cff51ac9
SK
561{
562 DeviceState *dev;
563 MachineState *ms = MACHINE(vms);
564 int irq = vms->irqmap[VIRT_ACPI_GED];
1962f31b 565 uint32_t event = ACPI_GED_PWR_DOWN_EVT;
cff51ac9
SK
566
567 if (ms->ram_slots) {
1962f31b 568 event |= ACPI_GED_MEM_HOTPLUG_EVT;
cff51ac9
SK
569 }
570
c2505d1c
SK
571 if (ms->nvdimms_state->is_enabled) {
572 event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
573 }
574
3e80f690 575 dev = qdev_new(TYPE_ACPI_GED);
cff51ac9
SK
576 qdev_prop_set_uint32(dev, "ged-event", event);
577
578 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
579 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
b8b69f4c 580 sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
cff51ac9 581
3c6ef471 582 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
cff51ac9
SK
583
584 return dev;
585}
586
b8b69f4c 587static void create_its(VirtMachineState *vms)
02f98731
PF
588{
589 const char *itsclass = its_class_name();
590 DeviceState *dev;
591
592 if (!itsclass) {
593 /* Do nothing if not supported */
594 return;
595 }
596
3e80f690 597 dev = qdev_new(itsclass);
02f98731 598
b8b69f4c 599 object_property_set_link(OBJECT(dev), OBJECT(vms->gic), "parent-gicv3",
02f98731 600 &error_abort);
3c6ef471 601 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
c8ef2bda 602 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
02f98731 603
c8ef2bda 604 fdt_add_its_gic_node(vms);
02f98731
PF
605}
606
b8b69f4c 607static void create_v2m(VirtMachineState *vms)
bd204e63
CD
608{
609 int i;
c8ef2bda 610 int irq = vms->irqmap[VIRT_GIC_V2M];
bd204e63
CD
611 DeviceState *dev;
612
3e80f690 613 dev = qdev_new("arm-gicv2m");
c8ef2bda 614 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
bd204e63
CD
615 qdev_prop_set_uint32(dev, "base-spi", irq);
616 qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
3c6ef471 617 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
bd204e63
CD
618
619 for (i = 0; i < NUM_GICV2M_SPIS; i++) {
b8b69f4c
PMD
620 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
621 qdev_get_gpio_in(vms->gic, irq + i));
bd204e63
CD
622 }
623
c8ef2bda 624 fdt_add_v2m_gic_node(vms);
bd204e63
CD
625}
626
b8b69f4c 627static void create_gic(VirtMachineState *vms)
64204743 628{
cc7d44c2 629 MachineState *ms = MACHINE(vms);
b92ad394 630 /* We create a standalone GIC */
64204743 631 SysBusDevice *gicbusdev;
e6fbcbc4 632 const char *gictype;
055a7f2b 633 int type = vms->gic_version, i;
cc7d44c2 634 unsigned int smp_cpus = ms->smp.cpus;
03d72fa1 635 uint32_t nb_redist_regions = 0;
64204743 636
b92ad394 637 gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
64204743 638
3e80f690 639 vms->gic = qdev_new(gictype);
b8b69f4c
PMD
640 qdev_prop_set_uint32(vms->gic, "revision", type);
641 qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
64204743
PM
642 /* Note that the num-irq property counts both internal and external
643 * interrupts; there are always 32 of the former (mandated by GIC spec).
644 */
b8b69f4c 645 qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
0e21f183 646 if (!kvm_irqchip_in_kernel()) {
b8b69f4c 647 qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
0e21f183 648 }
1e575b66
EA
649
650 if (type == 3) {
651 uint32_t redist0_capacity =
652 vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
653 uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
654
03d72fa1
EA
655 nb_redist_regions = virt_gicv3_redist_region_count(vms);
656
b8b69f4c 657 qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
03d72fa1 658 nb_redist_regions);
b8b69f4c 659 qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
03d72fa1
EA
660
661 if (nb_redist_regions == 2) {
662 uint32_t redist1_capacity =
bf424a12 663 vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
03d72fa1 664
b8b69f4c 665 qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
03d72fa1
EA
666 MIN(smp_cpus - redist0_count, redist1_capacity));
667 }
55ef3233
LM
668 } else {
669 if (!kvm_irqchip_in_kernel()) {
b8b69f4c 670 qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
55ef3233
LM
671 vms->virt);
672 }
1e575b66 673 }
b8b69f4c 674 gicbusdev = SYS_BUS_DEVICE(vms->gic);
3c6ef471 675 sysbus_realize_and_unref(gicbusdev, &error_fatal);
c8ef2bda 676 sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
b92ad394 677 if (type == 3) {
c8ef2bda 678 sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
03d72fa1 679 if (nb_redist_regions == 2) {
bf424a12
EA
680 sysbus_mmio_map(gicbusdev, 2,
681 vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
03d72fa1 682 }
b92ad394 683 } else {
c8ef2bda 684 sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
55ef3233
LM
685 if (vms->virt) {
686 sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
687 sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
688 }
b92ad394 689 }
64204743 690
5454006a
PM
691 /* Wire the outputs from each CPU's generic timer and the GICv3
692 * maintenance interrupt signal to the appropriate GIC PPI inputs,
693 * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
64204743
PM
694 */
695 for (i = 0; i < smp_cpus; i++) {
696 DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
0e3e858f 697 int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
a007b1f8
PM
698 int irq;
699 /* Mapping from the output timer irq lines from the CPU to the
700 * GIC PPI inputs we use for the virt board.
64204743 701 */
a007b1f8
PM
702 const int timer_irq[] = {
703 [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
704 [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
705 [GTIMER_HYP] = ARCH_TIMER_NS_EL2_IRQ,
706 [GTIMER_SEC] = ARCH_TIMER_S_EL1_IRQ,
707 };
708
709 for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
710 qdev_connect_gpio_out(cpudev, irq,
b8b69f4c 711 qdev_get_gpio_in(vms->gic,
a007b1f8
PM
712 ppibase + timer_irq[irq]));
713 }
64204743 714
55ef3233 715 if (type == 3) {
b8b69f4c 716 qemu_irq irq = qdev_get_gpio_in(vms->gic,
55ef3233
LM
717 ppibase + ARCH_GIC_MAINT_IRQ);
718 qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
719 0, irq);
720 } else if (vms->virt) {
b8b69f4c 721 qemu_irq irq = qdev_get_gpio_in(vms->gic,
55ef3233
LM
722 ppibase + ARCH_GIC_MAINT_IRQ);
723 sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
724 }
725
07f48730 726 qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
b8b69f4c 727 qdev_get_gpio_in(vms->gic, ppibase
07f48730 728 + VIRTUAL_PMU_IRQ));
5454006a 729
64204743 730 sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
8e7b4ca0
GB
731 sysbus_connect_irq(gicbusdev, i + smp_cpus,
732 qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
5454006a
PM
733 sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
734 qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
735 sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
736 qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
64204743
PM
737 }
738
055a7f2b 739 fdt_add_gic_node(vms);
bd204e63 740
ccc11b02 741 if (type == 3 && vms->its) {
b8b69f4c 742 create_its(vms);
2231f69b 743 } else if (type == 2) {
b8b69f4c 744 create_v2m(vms);
b92ad394 745 }
64204743
PM
746}
747
b8b69f4c 748static void create_uart(const VirtMachineState *vms, int uart,
0ec7b3e7 749 MemoryRegion *mem, Chardev *chr)
f5fdcd6e
PM
750{
751 char *nodename;
c8ef2bda
PM
752 hwaddr base = vms->memmap[uart].base;
753 hwaddr size = vms->memmap[uart].size;
754 int irq = vms->irqmap[uart];
f5fdcd6e
PM
755 const char compat[] = "arm,pl011\0arm,primecell";
756 const char clocknames[] = "uartclk\0apb_pclk";
3e80f690 757 DeviceState *dev = qdev_new(TYPE_PL011);
3df708eb 758 SysBusDevice *s = SYS_BUS_DEVICE(dev);
f5fdcd6e 759
9bbbf649 760 qdev_prop_set_chr(dev, "chardev", chr);
3c6ef471 761 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
3df708eb
PM
762 memory_region_add_subregion(mem, base,
763 sysbus_mmio_get_region(s, 0));
b8b69f4c 764 sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
f5fdcd6e
PM
765
766 nodename = g_strdup_printf("/pl011@%" PRIx64, base);
c8ef2bda 767 qemu_fdt_add_subnode(vms->fdt, nodename);
f5fdcd6e 768 /* Note that we can't use setprop_string because of the embedded NUL */
c8ef2bda 769 qemu_fdt_setprop(vms->fdt, nodename, "compatible",
f5fdcd6e 770 compat, sizeof(compat));
c8ef2bda 771 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
f5fdcd6e 772 2, base, 2, size);
c8ef2bda 773 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
f5fdcd6e 774 GIC_FDT_IRQ_TYPE_SPI, irq,
0be969a2 775 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
c8ef2bda
PM
776 qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
777 vms->clock_phandle, vms->clock_phandle);
778 qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
f5fdcd6e 779 clocknames, sizeof(clocknames));
f022b8e9 780
3df708eb 781 if (uart == VIRT_UART) {
c8ef2bda 782 qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
3df708eb
PM
783 } else {
784 /* Mark as not usable by the normal world */
c8ef2bda
PM
785 qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
786 qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
fb23d693 787
fb23d693
JF
788 qemu_fdt_setprop_string(vms->fdt, "/secure-chosen", "stdout-path",
789 nodename);
3df708eb
PM
790 }
791
f5fdcd6e
PM
792 g_free(nodename);
793}
794
b8b69f4c 795static void create_rtc(const VirtMachineState *vms)
6e411af9
PM
796{
797 char *nodename;
c8ef2bda
PM
798 hwaddr base = vms->memmap[VIRT_RTC].base;
799 hwaddr size = vms->memmap[VIRT_RTC].size;
800 int irq = vms->irqmap[VIRT_RTC];
6e411af9
PM
801 const char compat[] = "arm,pl031\0arm,primecell";
802
b8b69f4c 803 sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
6e411af9
PM
804
805 nodename = g_strdup_printf("/pl031@%" PRIx64, base);
c8ef2bda
PM
806 qemu_fdt_add_subnode(vms->fdt, nodename);
807 qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
808 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
6e411af9 809 2, base, 2, size);
c8ef2bda 810 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
6e411af9 811 GIC_FDT_IRQ_TYPE_SPI, irq,
0be969a2 812 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
c8ef2bda
PM
813 qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
814 qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
6e411af9
PM
815 g_free(nodename);
816}
817
94f02c5e 818static DeviceState *gpio_key_dev;
4bedd849
SZ
819static void virt_powerdown_req(Notifier *n, void *opaque)
820{
1962f31b
SK
821 VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
822
823 if (s->acpi_dev) {
824 acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
825 } else {
826 /* use gpio Pin 3 for power button event */
827 qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
828 }
4bedd849
SZ
829}
830
b8b69f4c 831static void create_gpio(const VirtMachineState *vms)
b0a3721e
SZ
832{
833 char *nodename;
94f02c5e 834 DeviceState *pl061_dev;
c8ef2bda
PM
835 hwaddr base = vms->memmap[VIRT_GPIO].base;
836 hwaddr size = vms->memmap[VIRT_GPIO].size;
837 int irq = vms->irqmap[VIRT_GPIO];
b0a3721e
SZ
838 const char compat[] = "arm,pl061\0arm,primecell";
839
b8b69f4c
PMD
840 pl061_dev = sysbus_create_simple("pl061", base,
841 qdev_get_gpio_in(vms->gic, irq));
b0a3721e 842
c8ef2bda 843 uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
b0a3721e 844 nodename = g_strdup_printf("/pl061@%" PRIx64, base);
c8ef2bda
PM
845 qemu_fdt_add_subnode(vms->fdt, nodename);
846 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
b0a3721e 847 2, base, 2, size);
c8ef2bda
PM
848 qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
849 qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
850 qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
851 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
b0a3721e
SZ
852 GIC_FDT_IRQ_TYPE_SPI, irq,
853 GIC_FDT_IRQ_FLAGS_LEVEL_HI);
c8ef2bda
PM
854 qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
855 qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
856 qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
3e6ebb64 857
94f02c5e
SZ
858 gpio_key_dev = sysbus_create_simple("gpio-key", -1,
859 qdev_get_gpio_in(pl061_dev, 3));
c8ef2bda
PM
860 qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
861 qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
862 qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
863 qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
3e6ebb64 864
c8ef2bda
PM
865 qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
866 qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
3e6ebb64 867 "label", "GPIO Key Poweroff");
c8ef2bda 868 qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
3e6ebb64 869 KEY_POWER);
c8ef2bda 870 qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
3e6ebb64 871 "gpios", phandle, 3, 0);
b0a3721e
SZ
872 g_free(nodename);
873}
874
b8b69f4c 875static void create_virtio_devices(const VirtMachineState *vms)
f5fdcd6e
PM
876{
877 int i;
c8ef2bda 878 hwaddr size = vms->memmap[VIRT_MMIO].size;
f5fdcd6e 879
587078f0
LE
880 /* We create the transports in forwards order. Since qbus_realize()
881 * prepends (not appends) new child buses, the incrementing loop below will
882 * create a list of virtio-mmio buses with decreasing base addresses.
883 *
884 * When a -device option is processed from the command line,
885 * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
886 * order. The upshot is that -device options in increasing command line
887 * order are mapped to virtio-mmio buses with decreasing base addresses.
888 *
889 * When this code was originally written, that arrangement ensured that the
890 * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
891 * the first -device on the command line. (The end-to-end order is a
892 * function of this loop, qbus_realize(), qbus_find_recursive(), and the
893 * guest kernel's name-to-address assignment strategy.)
894 *
895 * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
896 * the message, if not necessarily the code, of commit 70161ff336.
897 * Therefore the loop now establishes the inverse of the original intent.
898 *
899 * Unfortunately, we can't counteract the kernel change by reversing the
900 * loop; it would break existing command lines.
901 *
902 * In any case, the kernel makes no guarantee about the stability of
903 * enumeration order of virtio devices (as demonstrated by it changing
904 * between kernel versions). For reliable and stable identification
905 * of disks users must use UUIDs or similar mechanisms.
f5fdcd6e
PM
906 */
907 for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
c8ef2bda
PM
908 int irq = vms->irqmap[VIRT_MMIO] + i;
909 hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
f5fdcd6e 910
b8b69f4c
PMD
911 sysbus_create_simple("virtio-mmio", base,
912 qdev_get_gpio_in(vms->gic, irq));
f5fdcd6e
PM
913 }
914
587078f0
LE
915 /* We add dtb nodes in reverse order so that they appear in the finished
916 * device tree lowest address first.
917 *
918 * Note that this mapping is independent of the loop above. The previous
919 * loop influences virtio device to virtio transport assignment, whereas
920 * this loop controls how virtio transports are laid out in the dtb.
921 */
f5fdcd6e
PM
922 for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
923 char *nodename;
c8ef2bda
PM
924 int irq = vms->irqmap[VIRT_MMIO] + i;
925 hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
f5fdcd6e
PM
926
927 nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
c8ef2bda
PM
928 qemu_fdt_add_subnode(vms->fdt, nodename);
929 qemu_fdt_setprop_string(vms->fdt, nodename,
5a4348d1 930 "compatible", "virtio,mmio");
c8ef2bda 931 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
5a4348d1 932 2, base, 2, size);
c8ef2bda 933 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
5a4348d1
PC
934 GIC_FDT_IRQ_TYPE_SPI, irq,
935 GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
054bb7b2 936 qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
f5fdcd6e
PM
937 g_free(nodename);
938 }
939}
940
e0561e60
MA
941#define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
942
943static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
944 const char *name,
945 const char *alias_prop_name)
acf82361 946{
e0561e60
MA
947 /*
948 * Create a single flash device. We use the same parameters as
949 * the flash devices on the Versatile Express board.
acf82361 950 */
df707969 951 DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
acf82361 952
e0561e60 953 qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
acf82361
PM
954 qdev_prop_set_uint8(dev, "width", 4);
955 qdev_prop_set_uint8(dev, "device-width", 2);
e9809422 956 qdev_prop_set_bit(dev, "big-endian", false);
acf82361
PM
957 qdev_prop_set_uint16(dev, "id0", 0x89);
958 qdev_prop_set_uint16(dev, "id1", 0x18);
959 qdev_prop_set_uint16(dev, "id2", 0x00);
960 qdev_prop_set_uint16(dev, "id3", 0x00);
961 qdev_prop_set_string(dev, "name", name);
d2623129 962 object_property_add_child(OBJECT(vms), name, OBJECT(dev));
e0561e60 963 object_property_add_alias(OBJECT(vms), alias_prop_name,
d2623129 964 OBJECT(dev), "drive");
e0561e60
MA
965 return PFLASH_CFI01(dev);
966}
acf82361 967
e0561e60
MA
968static void virt_flash_create(VirtMachineState *vms)
969{
970 vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
971 vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
972}
acf82361 973
e0561e60
MA
974static void virt_flash_map1(PFlashCFI01 *flash,
975 hwaddr base, hwaddr size,
976 MemoryRegion *sysmem)
977{
978 DeviceState *dev = DEVICE(flash);
acf82361 979
4cdd0a77 980 assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
e0561e60
MA
981 assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
982 qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
3c6ef471 983 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
e0561e60
MA
984
985 memory_region_add_subregion(sysmem, base,
986 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
987 0));
16f4a8dc
PM
988}
989
e0561e60
MA
990static void virt_flash_map(VirtMachineState *vms,
991 MemoryRegion *sysmem,
992 MemoryRegion *secure_sysmem)
16f4a8dc 993{
e0561e60
MA
994 /*
995 * Map two flash devices to fill the VIRT_FLASH space in the memmap.
738a5d9f
PM
996 * sysmem is the system memory space. secure_sysmem is the secure view
997 * of the system, and the first flash device should be made visible only
998 * there. The second flash device is visible to both secure and nonsecure.
999 * If sysmem == secure_sysmem this means there is no separate Secure
1000 * address space and both flash devices are generally visible.
16f4a8dc 1001 */
c8ef2bda
PM
1002 hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1003 hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
acf82361 1004
e0561e60
MA
1005 virt_flash_map1(vms->flash[0], flashbase, flashsize,
1006 secure_sysmem);
1007 virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1008 sysmem);
1009}
1010
1011static void virt_flash_fdt(VirtMachineState *vms,
1012 MemoryRegion *sysmem,
1013 MemoryRegion *secure_sysmem)
1014{
1015 hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1016 hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1017 char *nodename;
acf82361 1018
738a5d9f
PM
1019 if (sysmem == secure_sysmem) {
1020 /* Report both flash devices as a single node in the DT */
1021 nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
c8ef2bda
PM
1022 qemu_fdt_add_subnode(vms->fdt, nodename);
1023 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1024 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
738a5d9f
PM
1025 2, flashbase, 2, flashsize,
1026 2, flashbase + flashsize, 2, flashsize);
c8ef2bda 1027 qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
738a5d9f
PM
1028 g_free(nodename);
1029 } else {
e0561e60
MA
1030 /*
1031 * Report the devices as separate nodes so we can mark one as
738a5d9f
PM
1032 * only visible to the secure world.
1033 */
1034 nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
c8ef2bda
PM
1035 qemu_fdt_add_subnode(vms->fdt, nodename);
1036 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1037 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
738a5d9f 1038 2, flashbase, 2, flashsize);
c8ef2bda
PM
1039 qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
1040 qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1041 qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
738a5d9f
PM
1042 g_free(nodename);
1043
1044 nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
c8ef2bda
PM
1045 qemu_fdt_add_subnode(vms->fdt, nodename);
1046 qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
1047 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
738a5d9f 1048 2, flashbase + flashsize, 2, flashsize);
c8ef2bda 1049 qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
738a5d9f
PM
1050 g_free(nodename);
1051 }
acf82361
PM
1052}
1053
e0561e60
MA
1054static bool virt_firmware_init(VirtMachineState *vms,
1055 MemoryRegion *sysmem,
1056 MemoryRegion *secure_sysmem)
1057{
1058 int i;
1059 BlockBackend *pflash_blk0;
1060
1061 /* Map legacy -drive if=pflash to machine properties */
1062 for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1063 pflash_cfi01_legacy_drive(vms->flash[i],
1064 drive_get(IF_PFLASH, 0, i));
1065 }
1066
1067 virt_flash_map(vms, sysmem, secure_sysmem);
1068
1069 pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1070
1071 if (bios_name) {
1072 char *fname;
1073 MemoryRegion *mr;
1074 int image_size;
1075
1076 if (pflash_blk0) {
1077 error_report("The contents of the first flash device may be "
1078 "specified with -bios or with -drive if=pflash... "
1079 "but you cannot use both options at once");
1080 exit(1);
1081 }
1082
1083 /* Fall back to -bios */
1084
1085 fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1086 if (!fname) {
1087 error_report("Could not find ROM image '%s'", bios_name);
1088 exit(1);
1089 }
1090 mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1091 image_size = load_image_mr(fname, mr);
1092 g_free(fname);
1093 if (image_size < 0) {
1094 error_report("Could not load ROM image '%s'", bios_name);
1095 exit(1);
1096 }
1097 }
1098
1099 return pflash_blk0 || bios_name;
1100}
1101
af1f60a4 1102static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
578f3c7b 1103{
cc7d44c2 1104 MachineState *ms = MACHINE(vms);
c8ef2bda
PM
1105 hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1106 hwaddr size = vms->memmap[VIRT_FW_CFG].size;
5836d168 1107 FWCfgState *fw_cfg;
578f3c7b
LE
1108 char *nodename;
1109
5836d168 1110 fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
cc7d44c2 1111 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
578f3c7b
LE
1112
1113 nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
c8ef2bda
PM
1114 qemu_fdt_add_subnode(vms->fdt, nodename);
1115 qemu_fdt_setprop_string(vms->fdt, nodename,
578f3c7b 1116 "compatible", "qemu,fw-cfg-mmio");
c8ef2bda 1117 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
578f3c7b 1118 2, base, 2, size);
14efdb5c 1119 qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
578f3c7b 1120 g_free(nodename);
af1f60a4 1121 return fw_cfg;
578f3c7b
LE
1122}
1123
c8ef2bda 1124static void create_pcie_irq_map(const VirtMachineState *vms,
9ac4ef77 1125 uint32_t gic_phandle,
4ab29b82
AG
1126 int first_irq, const char *nodename)
1127{
1128 int devfn, pin;
dfd90a87 1129 uint32_t full_irq_map[4 * 4 * 10] = { 0 };
4ab29b82
AG
1130 uint32_t *irq_map = full_irq_map;
1131
1132 for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1133 for (pin = 0; pin < 4; pin++) {
1134 int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1135 int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1136 int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1137 int i;
1138
1139 uint32_t map[] = {
1140 devfn << 8, 0, 0, /* devfn */
1141 pin + 1, /* PCI pin */
dfd90a87 1142 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
4ab29b82
AG
1143
1144 /* Convert map to big endian */
dfd90a87 1145 for (i = 0; i < 10; i++) {
4ab29b82
AG
1146 irq_map[i] = cpu_to_be32(map[i]);
1147 }
dfd90a87 1148 irq_map += 10;
4ab29b82
AG
1149 }
1150 }
1151
c8ef2bda 1152 qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
4ab29b82
AG
1153 full_irq_map, sizeof(full_irq_map));
1154
c8ef2bda 1155 qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
4ab29b82
AG
1156 0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
1157 0x7 /* PCI irq */);
1158}
1159
b8b69f4c 1160static void create_smmu(const VirtMachineState *vms,
584105ea
PM
1161 PCIBus *bus)
1162{
1163 char *node;
1164 const char compat[] = "arm,smmu-v3";
1165 int irq = vms->irqmap[VIRT_SMMU];
1166 int i;
1167 hwaddr base = vms->memmap[VIRT_SMMU].base;
1168 hwaddr size = vms->memmap[VIRT_SMMU].size;
1169 const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1170 DeviceState *dev;
1171
1172 if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1173 return;
1174 }
1175
3e80f690 1176 dev = qdev_new("arm-smmuv3");
584105ea
PM
1177
1178 object_property_set_link(OBJECT(dev), OBJECT(bus), "primary-bus",
1179 &error_abort);
3c6ef471 1180 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
584105ea
PM
1181 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1182 for (i = 0; i < NUM_SMMU_IRQS; i++) {
b8b69f4c
PMD
1183 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1184 qdev_get_gpio_in(vms->gic, irq + i));
584105ea
PM
1185 }
1186
1187 node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1188 qemu_fdt_add_subnode(vms->fdt, node);
1189 qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1190 qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg", 2, base, 2, size);
1191
1192 qemu_fdt_setprop_cells(vms->fdt, node, "interrupts",
1193 GIC_FDT_IRQ_TYPE_SPI, irq , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1194 GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1195 GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1196 GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1197
1198 qemu_fdt_setprop(vms->fdt, node, "interrupt-names", irq_names,
1199 sizeof(irq_names));
1200
1201 qemu_fdt_setprop_cell(vms->fdt, node, "clocks", vms->clock_phandle);
1202 qemu_fdt_setprop_string(vms->fdt, node, "clock-names", "apb_pclk");
1203 qemu_fdt_setprop(vms->fdt, node, "dma-coherent", NULL, 0);
1204
1205 qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1206
1207 qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1208 g_free(node);
1209}
1210
0fbddcec 1211static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
70e89132
EA
1212{
1213 const char compat[] = "virtio,pci-iommu";
1214 uint16_t bdf = vms->virtio_iommu_bdf;
1215 char *node;
1216
1217 vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1218
1219 node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1220 qemu_fdt_add_subnode(vms->fdt, node);
1221 qemu_fdt_setprop(vms->fdt, node, "compatible", compat, sizeof(compat));
1222 qemu_fdt_setprop_sized_cells(vms->fdt, node, "reg",
1223 1, bdf << 8, 1, 0, 1, 0,
1224 1, 0, 1, 0);
1225
1226 qemu_fdt_setprop_cell(vms->fdt, node, "#iommu-cells", 1);
1227 qemu_fdt_setprop_cell(vms->fdt, node, "phandle", vms->iommu_phandle);
1228 g_free(node);
1229
1230 qemu_fdt_setprop_cells(vms->fdt, vms->pciehb_nodename, "iommu-map",
1231 0x0, vms->iommu_phandle, 0x0, bdf,
1232 bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1233}
1234
b8b69f4c 1235static void create_pcie(VirtMachineState *vms)
4ab29b82 1236{
c8ef2bda
PM
1237 hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1238 hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
bf424a12
EA
1239 hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1240 hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
c8ef2bda
PM
1241 hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1242 hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
601d626d 1243 hwaddr base_ecam, size_ecam;
6a1f001b 1244 hwaddr base = base_mmio;
601d626d 1245 int nr_pcie_buses;
c8ef2bda 1246 int irq = vms->irqmap[VIRT_PCIE];
4ab29b82
AG
1247 MemoryRegion *mmio_alias;
1248 MemoryRegion *mmio_reg;
1249 MemoryRegion *ecam_alias;
1250 MemoryRegion *ecam_reg;
1251 DeviceState *dev;
1252 char *nodename;
601d626d 1253 int i, ecam_id;
fea9b3ca 1254 PCIHostState *pci;
4ab29b82 1255
3e80f690 1256 dev = qdev_new(TYPE_GPEX_HOST);
3c6ef471 1257 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
4ab29b82 1258
601d626d
EA
1259 ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1260 base_ecam = vms->memmap[ecam_id].base;
1261 size_ecam = vms->memmap[ecam_id].size;
1262 nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
4ab29b82
AG
1263 /* Map only the first size_ecam bytes of ECAM space */
1264 ecam_alias = g_new0(MemoryRegion, 1);
1265 ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1266 memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1267 ecam_reg, 0, size_ecam);
1268 memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1269
1270 /* Map the MMIO window into system address space so as to expose
1271 * the section of PCI MMIO space which starts at the same base address
1272 * (ie 1:1 mapping for that part of PCI MMIO space visible through
1273 * the window).
1274 */
1275 mmio_alias = g_new0(MemoryRegion, 1);
1276 mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1277 memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1278 mmio_reg, base_mmio, size_mmio);
1279 memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1280
0127937b 1281 if (vms->highmem) {
5125f9cd
PF
1282 /* Map high MMIO space */
1283 MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1284
1285 memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1286 mmio_reg, base_mmio_high, size_mmio_high);
1287 memory_region_add_subregion(get_system_memory(), base_mmio_high,
1288 high_mmio_alias);
1289 }
1290
4ab29b82 1291 /* Map IO port space */
6a1f001b 1292 sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
4ab29b82
AG
1293
1294 for (i = 0; i < GPEX_NUM_IRQS; i++) {
b8b69f4c
PMD
1295 sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1296 qdev_get_gpio_in(vms->gic, irq + i));
c9bb8e16 1297 gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
4ab29b82
AG
1298 }
1299
fea9b3ca
AK
1300 pci = PCI_HOST_BRIDGE(dev);
1301 if (pci->bus) {
1302 for (i = 0; i < nb_nics; i++) {
1303 NICInfo *nd = &nd_table[i];
1304
1305 if (!nd->model) {
1306 nd->model = g_strdup("virtio");
1307 }
1308
1309 pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1310 }
1311 }
1312
70e89132 1313 nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
c8ef2bda
PM
1314 qemu_fdt_add_subnode(vms->fdt, nodename);
1315 qemu_fdt_setprop_string(vms->fdt, nodename,
4ab29b82 1316 "compatible", "pci-host-ecam-generic");
c8ef2bda
PM
1317 qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1318 qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1319 qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
6d9c1b8d 1320 qemu_fdt_setprop_cell(vms->fdt, nodename, "linux,pci-domain", 0);
c8ef2bda 1321 qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
4ab29b82 1322 nr_pcie_buses - 1);
c8ef2bda 1323 qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
4ab29b82 1324
c8ef2bda
PM
1325 if (vms->msi_phandle) {
1326 qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1327 vms->msi_phandle);
b92ad394 1328 }
bd204e63 1329
c8ef2bda 1330 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
4ab29b82 1331 2, base_ecam, 2, size_ecam);
5125f9cd 1332
0127937b 1333 if (vms->highmem) {
c8ef2bda 1334 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
5125f9cd
PF
1335 1, FDT_PCI_RANGE_IOPORT, 2, 0,
1336 2, base_pio, 2, size_pio,
1337 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1338 2, base_mmio, 2, size_mmio,
1339 1, FDT_PCI_RANGE_MMIO_64BIT,
1340 2, base_mmio_high,
1341 2, base_mmio_high, 2, size_mmio_high);
1342 } else {
c8ef2bda 1343 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
5125f9cd
PF
1344 1, FDT_PCI_RANGE_IOPORT, 2, 0,
1345 2, base_pio, 2, size_pio,
1346 1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1347 2, base_mmio, 2, size_mmio);
1348 }
4ab29b82 1349
c8ef2bda
PM
1350 qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1351 create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
4ab29b82 1352
584105ea
PM
1353 if (vms->iommu) {
1354 vms->iommu_phandle = qemu_fdt_alloc_phandle(vms->fdt);
1355
70e89132
EA
1356 switch (vms->iommu) {
1357 case VIRT_IOMMU_SMMUV3:
1358 create_smmu(vms, pci->bus);
1359 qemu_fdt_setprop_cells(vms->fdt, nodename, "iommu-map",
1360 0x0, vms->iommu_phandle, 0x0, 0x10000);
1361 break;
1362 default:
1363 g_assert_not_reached();
1364 }
584105ea 1365 }
4ab29b82
AG
1366}
1367
b8b69f4c 1368static void create_platform_bus(VirtMachineState *vms)
5f7a5a0e
EA
1369{
1370 DeviceState *dev;
1371 SysBusDevice *s;
1372 int i;
5f7a5a0e
EA
1373 MemoryRegion *sysmem = get_system_memory();
1374
3e80f690 1375 dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
5f7a5a0e 1376 dev->id = TYPE_PLATFORM_BUS_DEVICE;
3b77f6c3
IM
1377 qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1378 qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
3c6ef471 1379 sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
a3fc8396 1380 vms->platform_bus_dev = dev;
5f7a5a0e 1381
3b77f6c3
IM
1382 s = SYS_BUS_DEVICE(dev);
1383 for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
b8b69f4c
PMD
1384 int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1385 sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
5f7a5a0e
EA
1386 }
1387
1388 memory_region_add_subregion(sysmem,
3b77f6c3 1389 vms->memmap[VIRT_PLATFORM_BUS].base,
5f7a5a0e
EA
1390 sysbus_mmio_get_region(s, 0));
1391}
1392
c8ef2bda 1393static void create_secure_ram(VirtMachineState *vms,
9ac4ef77 1394 MemoryRegion *secure_sysmem)
83ec1923
PM
1395{
1396 MemoryRegion *secram = g_new(MemoryRegion, 1);
1397 char *nodename;
c8ef2bda
PM
1398 hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1399 hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
83ec1923 1400
98a99ce0
PM
1401 memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1402 &error_fatal);
83ec1923
PM
1403 memory_region_add_subregion(secure_sysmem, base, secram);
1404
1405 nodename = g_strdup_printf("/secram@%" PRIx64, base);
c8ef2bda
PM
1406 qemu_fdt_add_subnode(vms->fdt, nodename);
1407 qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1408 qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1409 qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1410 qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
83ec1923
PM
1411
1412 g_free(nodename);
1413}
1414
f5fdcd6e
PM
1415static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1416{
9ac4ef77
PM
1417 const VirtMachineState *board = container_of(binfo, VirtMachineState,
1418 bootinfo);
f5fdcd6e
PM
1419
1420 *fdt_size = board->fdt_size;
1421 return board->fdt;
1422}
1423
e9a8e474 1424static void virt_build_smbios(VirtMachineState *vms)
c30e1565 1425{
dfadc3bf
WH
1426 MachineClass *mc = MACHINE_GET_CLASS(vms);
1427 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
c30e1565
WH
1428 uint8_t *smbios_tables, *smbios_anchor;
1429 size_t smbios_tables_len, smbios_anchor_len;
bab27ea2 1430 const char *product = "QEMU Virtual Machine";
c30e1565 1431
bab27ea2
AJ
1432 if (kvm_enabled()) {
1433 product = "KVM Virtual Machine";
1434 }
1435
1436 smbios_set_defaults("QEMU", product,
dfadc3bf
WH
1437 vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1438 true, SMBIOS_ENTRY_POINT_30);
c30e1565 1439
a0628599 1440 smbios_get_tables(MACHINE(vms), NULL, 0, &smbios_tables, &smbios_tables_len,
c30e1565
WH
1441 &smbios_anchor, &smbios_anchor_len);
1442
1443 if (smbios_anchor) {
af1f60a4 1444 fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
c30e1565 1445 smbios_tables, smbios_tables_len);
af1f60a4 1446 fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
c30e1565
WH
1447 smbios_anchor, smbios_anchor_len);
1448 }
1449}
1450
d7c2e2db 1451static
054f4dc9 1452void virt_machine_done(Notifier *notifier, void *data)
d7c2e2db 1453{
054f4dc9
AJ
1454 VirtMachineState *vms = container_of(notifier, VirtMachineState,
1455 machine_done);
2744ece8 1456 MachineState *ms = MACHINE(vms);
3b77f6c3
IM
1457 ARMCPU *cpu = ARM_CPU(first_cpu);
1458 struct arm_boot_info *info = &vms->bootinfo;
1459 AddressSpace *as = arm_boot_address_space(cpu, info);
1460
1461 /*
1462 * If the user provided a dtb, we assume the dynamic sysbus nodes
1463 * already are integrated there. This corresponds to a use case where
1464 * the dynamic sysbus nodes are complex and their generation is not yet
1465 * supported. In that case the user can take charge of the guest dt
1466 * while qemu takes charge of the qom stuff.
1467 */
1468 if (info->dtb_filename == NULL) {
1469 platform_bus_add_all_fdt_nodes(vms->fdt, "/intc",
1470 vms->memmap[VIRT_PLATFORM_BUS].base,
1471 vms->memmap[VIRT_PLATFORM_BUS].size,
1472 vms->irqmap[VIRT_PLATFORM_BUS]);
1473 }
2744ece8 1474 if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
3b77f6c3
IM
1475 exit(1);
1476 }
054f4dc9 1477
e9a8e474
AJ
1478 virt_acpi_setup(vms);
1479 virt_build_smbios(vms);
d7c2e2db
SZ
1480}
1481
46de5913
IM
1482static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1483{
1484 uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1485 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1486
1487 if (!vmc->disallow_affinity_adjustment) {
1488 /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1489 * GIC's target-list limitations. 32-bit KVM hosts currently
1490 * always create clusters of 4 CPUs, but that is expected to
1491 * change when they gain support for gicv3. When KVM is enabled
1492 * it will override the changes we make here, therefore our
1493 * purposes are to make TCG consistent (with 64-bit KVM hosts)
1494 * and to improve SGI efficiency.
1495 */
d04460e5 1496 if (vms->gic_version == VIRT_GIC_VERSION_3) {
46de5913
IM
1497 clustersz = GICV3_TARGETLIST_BITS;
1498 } else {
1499 clustersz = GIC_TARGETLIST_BITS;
1500 }
1501 }
1502 return arm_cpu_mp_affinity(idx, clustersz);
1503}
1504
350a9c9e
EA
1505static void virt_set_memmap(VirtMachineState *vms)
1506{
957e32cf
EA
1507 MachineState *ms = MACHINE(vms);
1508 hwaddr base, device_memory_base, device_memory_size;
350a9c9e
EA
1509 int i;
1510
1511 vms->memmap = extended_memmap;
1512
1513 for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1514 vms->memmap[i] = base_memmap[i];
1515 }
1516
957e32cf
EA
1517 if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1518 error_report("unsupported number of memory slots: %"PRIu64,
1519 ms->ram_slots);
1520 exit(EXIT_FAILURE);
1521 }
1522
1523 /*
1524 * We compute the base of the high IO region depending on the
1525 * amount of initial and device memory. The device memory start/size
1526 * is aligned on 1GiB. We never put the high IO region below 256GiB
1527 * so that if maxram_size is < 255GiB we keep the legacy memory map.
1528 * The device region size assumes 1GiB page max alignment per slot.
1529 */
1530 device_memory_base =
1531 ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1532 device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1533
1534 /* Base address of the high IO region */
1535 base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1536 if (base < device_memory_base) {
1537 error_report("maxmem/slots too huge");
1538 exit(EXIT_FAILURE);
1539 }
1540 if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1541 base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1542 }
350a9c9e
EA
1543
1544 for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1545 hwaddr size = extended_memmap[i].size;
1546
1547 base = ROUND_UP(base, size);
1548 vms->memmap[i].base = base;
1549 vms->memmap[i].size = size;
1550 base += size;
1551 }
957e32cf
EA
1552 vms->highest_gpa = base - 1;
1553 if (device_memory_size > 0) {
1554 ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1555 ms->device_memory->base = device_memory_base;
1556 memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1557 "device-memory", device_memory_size);
1558 }
350a9c9e
EA
1559}
1560
36bf4ec8
EA
1561/*
1562 * finalize_gic_version - Determines the final gic_version
1563 * according to the gic-version property
1564 *
1565 * Default GIC type is v2
1566 */
1567static void finalize_gic_version(VirtMachineState *vms)
1568{
6785aee0
EA
1569 unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1570
97b4c918
EA
1571 if (kvm_enabled()) {
1572 int probe_bitmap;
d45efe47 1573
97b4c918
EA
1574 if (!kvm_irqchip_in_kernel()) {
1575 switch (vms->gic_version) {
1576 case VIRT_GIC_VERSION_HOST:
1577 warn_report(
1578 "gic-version=host not relevant with kernel-irqchip=off "
1579 "as only userspace GICv2 is supported. Using v2 ...");
1580 return;
1581 case VIRT_GIC_VERSION_MAX:
1582 case VIRT_GIC_VERSION_NOSEL:
1583 vms->gic_version = VIRT_GIC_VERSION_2;
1584 return;
1585 case VIRT_GIC_VERSION_2:
1586 return;
1587 case VIRT_GIC_VERSION_3:
36bf4ec8 1588 error_report(
97b4c918 1589 "gic-version=3 is not supported with kernel-irqchip=off");
36bf4ec8 1590 exit(1);
97b4c918
EA
1591 }
1592 }
1593
1594 probe_bitmap = kvm_arm_vgic_probe();
1595 if (!probe_bitmap) {
1596 error_report("Unable to determine GIC version supported by host");
1597 exit(1);
1598 }
1599
1600 switch (vms->gic_version) {
1601 case VIRT_GIC_VERSION_HOST:
1602 case VIRT_GIC_VERSION_MAX:
1603 if (probe_bitmap & KVM_ARM_VGIC_V3) {
1604 vms->gic_version = VIRT_GIC_VERSION_3;
d45efe47 1605 } else {
97b4c918 1606 vms->gic_version = VIRT_GIC_VERSION_2;
36bf4ec8 1607 }
97b4c918
EA
1608 return;
1609 case VIRT_GIC_VERSION_NOSEL:
6785aee0
EA
1610 if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1611 vms->gic_version = VIRT_GIC_VERSION_2;
1612 } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1613 /*
1614 * in case the host does not support v2 in-kernel emulation or
1615 * the end-user requested more than 8 VCPUs we now default
1616 * to v3. In any case defaulting to v2 would be broken.
1617 */
1618 vms->gic_version = VIRT_GIC_VERSION_3;
1619 } else if (max_cpus > GIC_NCPU) {
1620 error_report("host only supports in-kernel GICv2 emulation "
1621 "but more than 8 vcpus are requested");
1622 exit(1);
1623 }
97b4c918
EA
1624 break;
1625 case VIRT_GIC_VERSION_2:
1626 case VIRT_GIC_VERSION_3:
1627 break;
1628 }
1629
1630 /* Check chosen version is effectively supported by the host */
1631 if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1632 !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1633 error_report("host does not support in-kernel GICv2 emulation");
1634 exit(1);
1635 } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1636 !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1637 error_report("host does not support in-kernel GICv3 emulation");
1638 exit(1);
36bf4ec8 1639 }
97b4c918
EA
1640 return;
1641 }
1642
1643 /* TCG mode */
1644 switch (vms->gic_version) {
1645 case VIRT_GIC_VERSION_NOSEL:
36bf4ec8 1646 vms->gic_version = VIRT_GIC_VERSION_2;
97b4c918
EA
1647 break;
1648 case VIRT_GIC_VERSION_MAX:
1649 vms->gic_version = VIRT_GIC_VERSION_3;
1650 break;
1651 case VIRT_GIC_VERSION_HOST:
1652 error_report("gic-version=host requires KVM");
1653 exit(1);
1654 case VIRT_GIC_VERSION_2:
1655 case VIRT_GIC_VERSION_3:
1656 break;
36bf4ec8
EA
1657 }
1658}
1659
3ef96221 1660static void machvirt_init(MachineState *machine)
f5fdcd6e 1661{
e5a5604f 1662 VirtMachineState *vms = VIRT_MACHINE(machine);
95eb49c8 1663 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
17d3d0e2
IM
1664 MachineClass *mc = MACHINE_GET_CLASS(machine);
1665 const CPUArchIdList *possible_cpus;
f5fdcd6e 1666 MemoryRegion *sysmem = get_system_memory();
3df708eb 1667 MemoryRegion *secure_sysmem = NULL;
7ea686f5 1668 int n, virt_max_cpus;
e0561e60 1669 bool firmware_loaded;
17ec075a 1670 bool aarch64 = true;
cff51ac9 1671 bool has_ged = !vmc->no_ged;
cc7d44c2
LX
1672 unsigned int smp_cpus = machine->smp.cpus;
1673 unsigned int max_cpus = machine->smp.max_cpus;
f5fdcd6e 1674
c9650222
EA
1675 /*
1676 * In accelerated mode, the memory map is computed earlier in kvm_type()
1677 * to create a VM with the right number of IPA bits.
1678 */
1679 if (!vms->memmap) {
1680 virt_set_memmap(vms);
1681 }
350a9c9e 1682
b92ad394
PF
1683 /* We can probe only here because during property set
1684 * KVM is not available yet
1685 */
36bf4ec8 1686 finalize_gic_version(vms);
b92ad394 1687
ba1ba5cc
IM
1688 if (!cpu_type_valid(machine->cpu_type)) {
1689 error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
f5fdcd6e
PM
1690 exit(1);
1691 }
1692
e0561e60
MA
1693 if (vms->secure) {
1694 if (kvm_enabled()) {
1695 error_report("mach-virt: KVM does not support Security extensions");
1696 exit(1);
1697 }
1698
1699 /*
1700 * The Secure view of the world is the same as the NonSecure,
1701 * but with a few extra devices. Create it as a container region
1702 * containing the system memory at low priority; any secure-only
1703 * devices go in at higher priority and take precedence.
1704 */
1705 secure_sysmem = g_new(MemoryRegion, 1);
1706 memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1707 UINT64_MAX);
1708 memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1709 }
1710
1711 firmware_loaded = virt_firmware_init(vms, sysmem,
1712 secure_sysmem ?: sysmem);
1713
4824a61a
PM
1714 /* If we have an EL3 boot ROM then the assumption is that it will
1715 * implement PSCI itself, so disable QEMU's internal implementation
1716 * so it doesn't get in the way. Instead of starting secondary
1717 * CPUs in PSCI powerdown state we will start them all running and
1718 * let the boot ROM sort them out.
f29cacfb
PM
1719 * The usual case is that we do use QEMU's PSCI implementation;
1720 * if the guest has EL2 then we will use SMC as the conduit,
1721 * and otherwise we will use HVC (for backwards compatibility and
1722 * because if we're using KVM then we must use HVC).
4824a61a 1723 */
2013c566
PM
1724 if (vms->secure && firmware_loaded) {
1725 vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
f29cacfb
PM
1726 } else if (vms->virt) {
1727 vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2013c566
PM
1728 } else {
1729 vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1730 }
4824a61a 1731
4b280b72
AJ
1732 /* The maximum number of CPUs depends on the GIC version, or on how
1733 * many redistributors we can fit into the memory map.
1734 */
d04460e5 1735 if (vms->gic_version == VIRT_GIC_VERSION_3) {
bf424a12
EA
1736 virt_max_cpus =
1737 vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1738 virt_max_cpus +=
1739 vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
4b280b72 1740 } else {
7ea686f5 1741 virt_max_cpus = GIC_NCPU;
4b280b72
AJ
1742 }
1743
7ea686f5 1744 if (max_cpus > virt_max_cpus) {
4b280b72
AJ
1745 error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1746 "supported by machine 'mach-virt' (%d)",
7ea686f5 1747 max_cpus, virt_max_cpus);
4b280b72
AJ
1748 exit(1);
1749 }
1750
c8ef2bda 1751 vms->smp_cpus = smp_cpus;
f5fdcd6e 1752
f29cacfb
PM
1753 if (vms->virt && kvm_enabled()) {
1754 error_report("mach-virt: KVM does not support providing "
1755 "Virtualization extensions to the guest CPU");
1756 exit(1);
1757 }
1758
c8ef2bda 1759 create_fdt(vms);
f5fdcd6e 1760
17d3d0e2
IM
1761 possible_cpus = mc->possible_cpu_arch_ids(machine);
1762 for (n = 0; n < possible_cpus->len; n++) {
1763 Object *cpuobj;
d9c34f9c 1764 CPUState *cs;
46de5913 1765
17d3d0e2
IM
1766 if (n >= smp_cpus) {
1767 break;
1768 }
1769
d342eb76 1770 cpuobj = object_new(possible_cpus->cpus[n].type);
17d3d0e2 1771 object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
46de5913 1772 "mp-affinity", NULL);
f313369f 1773
d9c34f9c
IM
1774 cs = CPU(cpuobj);
1775 cs->cpu_index = n;
1776
a0ceb640
IM
1777 numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1778 &error_fatal);
bd4c1bfe 1779
17ec075a
EA
1780 aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1781
e5a5604f
GB
1782 if (!vms->secure) {
1783 object_property_set_bool(cpuobj, false, "has_el3", NULL);
1784 }
1785
f29cacfb 1786 if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
c25bd18a
PM
1787 object_property_set_bool(cpuobj, false, "has_el2", NULL);
1788 }
1789
2013c566
PM
1790 if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1791 object_property_set_int(cpuobj, vms->psci_conduit,
4824a61a 1792 "psci-conduit", NULL);
211b0169 1793
4824a61a
PM
1794 /* Secondary CPUs start in PSCI powered-down state */
1795 if (n > 0) {
1796 object_property_set_bool(cpuobj, true,
1797 "start-powered-off", NULL);
1798 }
f5fdcd6e 1799 }
ba750085 1800
dea101a1
AJ
1801 if (vmc->kvm_no_adjvtime &&
1802 object_property_find(cpuobj, "kvm-no-adjvtime", NULL)) {
1803 object_property_set_bool(cpuobj, true, "kvm-no-adjvtime", NULL);
1804 }
1805
1141d1eb
WH
1806 if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1807 object_property_set_bool(cpuobj, false, "pmu", NULL);
1808 }
1809
ba750085 1810 if (object_property_find(cpuobj, "reset-cbar", NULL)) {
c8ef2bda 1811 object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
ba750085
PM
1812 "reset-cbar", &error_abort);
1813 }
1814
1d939a68
PM
1815 object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1816 &error_abort);
3df708eb
PM
1817 if (vms->secure) {
1818 object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1819 "secure-memory", &error_abort);
1820 }
1d939a68 1821
ce189ab2 1822 qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
dbb74759 1823 object_unref(cpuobj);
f5fdcd6e 1824 }
055a7f2b 1825 fdt_add_timer_nodes(vms);
c8ef2bda 1826 fdt_add_cpu_nodes(vms);
f5fdcd6e 1827
2ba956cc
EA
1828 if (!kvm_enabled()) {
1829 ARMCPU *cpu = ARM_CPU(first_cpu);
1830 bool aarch64 = object_property_get_bool(OBJECT(cpu), "aarch64", NULL);
1831
1832 if (aarch64 && vms->highmem) {
1833 int requested_pa_size, pamax = arm_pamax(cpu);
1834
1835 requested_pa_size = 64 - clz64(vms->highest_gpa);
1836 if (pamax < requested_pa_size) {
1837 error_report("VCPU supports less PA bits (%d) than requested "
1838 "by the memory map (%d)", pamax, requested_pa_size);
1839 exit(1);
1840 }
1841 }
1842 }
1843
a72f6805
IM
1844 memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
1845 machine->ram);
957e32cf
EA
1846 if (machine->device_memory) {
1847 memory_region_add_subregion(sysmem, machine->device_memory->base,
1848 &machine->device_memory->mr);
1849 }
f5fdcd6e 1850
80734cbd 1851 virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
acf82361 1852
b8b69f4c 1853 create_gic(vms);
f5fdcd6e 1854
055a7f2b 1855 fdt_add_pmu_nodes(vms);
01fe6b60 1856
b8b69f4c 1857 create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
3df708eb
PM
1858
1859 if (vms->secure) {
c8ef2bda 1860 create_secure_ram(vms, secure_sysmem);
b8b69f4c 1861 create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
3df708eb 1862 }
f5fdcd6e 1863
17ec075a
EA
1864 vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
1865
b8b69f4c 1866 create_rtc(vms);
6e411af9 1867
b8b69f4c 1868 create_pcie(vms);
4ab29b82 1869
17e89077 1870 if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
b8b69f4c 1871 vms->acpi_dev = create_acpi_ged(vms);
1962f31b 1872 } else {
b8b69f4c 1873 create_gpio(vms);
cff51ac9
SK
1874 }
1875
c345680c
SK
1876 /* connect powerdown request */
1877 vms->powerdown_notifier.notify = virt_powerdown_req;
1878 qemu_register_powerdown_notifier(&vms->powerdown_notifier);
1879
f5fdcd6e
PM
1880 /* Create mmio transports, so the user can create virtio backends
1881 * (which will be automatically plugged in to the transports). If
1882 * no backend is created the transport will just sit harmlessly idle.
1883 */
b8b69f4c 1884 create_virtio_devices(vms);
f5fdcd6e 1885
af1f60a4
AJ
1886 vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1887 rom_set_fw(vms->fw_cfg);
d7c2e2db 1888
b8b69f4c 1889 create_platform_bus(vms);
578f3c7b 1890
b5a60bee
KL
1891 if (machine->nvdimms_state->is_enabled) {
1892 const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
1893 .space_id = AML_AS_SYSTEM_MEMORY,
1894 .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
1895 .bit_width = NVDIMM_ACPI_IO_LEN << 3
1896 };
1897
1898 nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
1899 arm_virt_nvdimm_acpi_dsmio,
1900 vms->fw_cfg, OBJECT(vms));
1901 }
1902
c8ef2bda 1903 vms->bootinfo.ram_size = machine->ram_size;
c8ef2bda
PM
1904 vms->bootinfo.nb_cpus = smp_cpus;
1905 vms->bootinfo.board_id = -1;
1906 vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1907 vms->bootinfo.get_dtb = machvirt_dtb;
3b77f6c3 1908 vms->bootinfo.skip_dtb_autoload = true;
c8ef2bda 1909 vms->bootinfo.firmware_loaded = firmware_loaded;
2744ece8 1910 arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
5f7a5a0e 1911
3b77f6c3
IM
1912 vms->machine_done.notify = virt_machine_done;
1913 qemu_add_machine_init_done_notifier(&vms->machine_done);
f5fdcd6e
PM
1914}
1915
083a5890
GB
1916static bool virt_get_secure(Object *obj, Error **errp)
1917{
1918 VirtMachineState *vms = VIRT_MACHINE(obj);
1919
1920 return vms->secure;
1921}
1922
1923static void virt_set_secure(Object *obj, bool value, Error **errp)
1924{
1925 VirtMachineState *vms = VIRT_MACHINE(obj);
1926
1927 vms->secure = value;
1928}
1929
f29cacfb
PM
1930static bool virt_get_virt(Object *obj, Error **errp)
1931{
1932 VirtMachineState *vms = VIRT_MACHINE(obj);
1933
1934 return vms->virt;
1935}
1936
1937static void virt_set_virt(Object *obj, bool value, Error **errp)
1938{
1939 VirtMachineState *vms = VIRT_MACHINE(obj);
1940
1941 vms->virt = value;
1942}
1943
5125f9cd
PF
1944static bool virt_get_highmem(Object *obj, Error **errp)
1945{
1946 VirtMachineState *vms = VIRT_MACHINE(obj);
1947
1948 return vms->highmem;
1949}
1950
1951static void virt_set_highmem(Object *obj, bool value, Error **errp)
1952{
1953 VirtMachineState *vms = VIRT_MACHINE(obj);
1954
1955 vms->highmem = value;
1956}
1957
ccc11b02
EA
1958static bool virt_get_its(Object *obj, Error **errp)
1959{
1960 VirtMachineState *vms = VIRT_MACHINE(obj);
1961
1962 return vms->its;
1963}
1964
1965static void virt_set_its(Object *obj, bool value, Error **errp)
1966{
1967 VirtMachineState *vms = VIRT_MACHINE(obj);
1968
1969 vms->its = value;
1970}
1971
17e89077
GH
1972bool virt_is_acpi_enabled(VirtMachineState *vms)
1973{
1974 if (vms->acpi == ON_OFF_AUTO_OFF) {
1975 return false;
1976 }
1977 return true;
1978}
1979
1980static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
1981 void *opaque, Error **errp)
1982{
1983 VirtMachineState *vms = VIRT_MACHINE(obj);
1984 OnOffAuto acpi = vms->acpi;
1985
1986 visit_type_OnOffAuto(v, name, &acpi, errp);
1987}
1988
1989static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
1990 void *opaque, Error **errp)
1991{
1992 VirtMachineState *vms = VIRT_MACHINE(obj);
1993
1994 visit_type_OnOffAuto(v, name, &vms->acpi, errp);
1995}
1996
2afa8c85
DG
1997static bool virt_get_ras(Object *obj, Error **errp)
1998{
1999 VirtMachineState *vms = VIRT_MACHINE(obj);
2000
2001 return vms->ras;
2002}
2003
2004static void virt_set_ras(Object *obj, bool value, Error **errp)
2005{
2006 VirtMachineState *vms = VIRT_MACHINE(obj);
2007
2008 vms->ras = value;
2009}
2010
b92ad394
PF
2011static char *virt_get_gic_version(Object *obj, Error **errp)
2012{
2013 VirtMachineState *vms = VIRT_MACHINE(obj);
d04460e5 2014 const char *val = vms->gic_version == VIRT_GIC_VERSION_3 ? "3" : "2";
b92ad394
PF
2015
2016 return g_strdup(val);
2017}
2018
2019static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2020{
2021 VirtMachineState *vms = VIRT_MACHINE(obj);
2022
2023 if (!strcmp(value, "3")) {
d04460e5 2024 vms->gic_version = VIRT_GIC_VERSION_3;
b92ad394 2025 } else if (!strcmp(value, "2")) {
d04460e5 2026 vms->gic_version = VIRT_GIC_VERSION_2;
b92ad394 2027 } else if (!strcmp(value, "host")) {
d04460e5 2028 vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
dc16538a 2029 } else if (!strcmp(value, "max")) {
d04460e5 2030 vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
b92ad394 2031 } else {
7b55044f 2032 error_setg(errp, "Invalid gic-version value");
dc16538a 2033 error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
b92ad394
PF
2034 }
2035}
2036
e24e3454
EA
2037static char *virt_get_iommu(Object *obj, Error **errp)
2038{
2039 VirtMachineState *vms = VIRT_MACHINE(obj);
2040
2041 switch (vms->iommu) {
2042 case VIRT_IOMMU_NONE:
2043 return g_strdup("none");
2044 case VIRT_IOMMU_SMMUV3:
2045 return g_strdup("smmuv3");
2046 default:
2047 g_assert_not_reached();
2048 }
2049}
2050
2051static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2052{
2053 VirtMachineState *vms = VIRT_MACHINE(obj);
2054
2055 if (!strcmp(value, "smmuv3")) {
2056 vms->iommu = VIRT_IOMMU_SMMUV3;
2057 } else if (!strcmp(value, "none")) {
2058 vms->iommu = VIRT_IOMMU_NONE;
2059 } else {
2060 error_setg(errp, "Invalid iommu value");
2061 error_append_hint(errp, "Valid values are none, smmuv3.\n");
2062 }
2063}
2064
ea089eeb
IM
2065static CpuInstanceProperties
2066virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2067{
2068 MachineClass *mc = MACHINE_GET_CLASS(ms);
2069 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2070
2071 assert(cpu_index < possible_cpus->len);
2072 return possible_cpus->cpus[cpu_index].props;
2073}
2074
79e07936
IM
2075static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2076{
aa570207 2077 return idx % ms->numa_state->num_nodes;
79e07936
IM
2078}
2079
17d3d0e2
IM
2080static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2081{
2082 int n;
cc7d44c2 2083 unsigned int max_cpus = ms->smp.max_cpus;
17d3d0e2
IM
2084 VirtMachineState *vms = VIRT_MACHINE(ms);
2085
2086 if (ms->possible_cpus) {
2087 assert(ms->possible_cpus->len == max_cpus);
2088 return ms->possible_cpus;
2089 }
2090
2091 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2092 sizeof(CPUArchId) * max_cpus);
2093 ms->possible_cpus->len = max_cpus;
2094 for (n = 0; n < ms->possible_cpus->len; n++) {
d342eb76 2095 ms->possible_cpus->cpus[n].type = ms->cpu_type;
17d3d0e2
IM
2096 ms->possible_cpus->cpus[n].arch_id =
2097 virt_cpu_mp_affinity(vms, n);
2098 ms->possible_cpus->cpus[n].props.has_thread_id = true;
2099 ms->possible_cpus->cpus[n].props.thread_id = n;
17d3d0e2
IM
2100 }
2101 return ms->possible_cpus;
2102}
2103
1f283ae1
EA
2104static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2105 Error **errp)
2106{
cff51ac9 2107 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
c2505d1c 2108 const MachineState *ms = MACHINE(hotplug_dev);
cff51ac9 2109 const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1f283ae1 2110
cff51ac9
SK
2111 if (!vms->acpi_dev) {
2112 error_setg(errp,
2113 "memory hotplug is not enabled: missing acpi-ged device");
1f283ae1
EA
2114 return;
2115 }
2116
c2505d1c
SK
2117 if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2118 error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2119 return;
2120 }
2121
1f283ae1
EA
2122 pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2123}
2124
2125static void virt_memory_plug(HotplugHandler *hotplug_dev,
2126 DeviceState *dev, Error **errp)
2127{
2128 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
b5a60bee
KL
2129 MachineState *ms = MACHINE(hotplug_dev);
2130 bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1f283ae1
EA
2131 Error *local_err = NULL;
2132
2133 pc_dimm_plug(PC_DIMM(dev), MACHINE(vms), &local_err);
cff51ac9
SK
2134 if (local_err) {
2135 goto out;
2136 }
1f283ae1 2137
b5a60bee
KL
2138 if (is_nvdimm) {
2139 nvdimm_plug(ms->nvdimms_state);
2140 }
2141
53eccc70
KZ
2142 hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2143 dev, &error_abort);
2144
cff51ac9 2145out:
1f283ae1
EA
2146 error_propagate(errp, local_err);
2147}
2148
2149static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2150 DeviceState *dev, Error **errp)
2151{
2152 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2153 virt_memory_pre_plug(hotplug_dev, dev, errp);
2154 }
2155}
2156
a3fc8396
IM
2157static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2158 DeviceState *dev, Error **errp)
2159{
2160 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2161
2162 if (vms->platform_bus_dev) {
2163 if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE)) {
2164 platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2165 SYS_BUS_DEVICE(dev));
2166 }
2167 }
1f283ae1
EA
2168 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2169 virt_memory_plug(hotplug_dev, dev, errp);
2170 }
70e89132
EA
2171 if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2172 PCIDevice *pdev = PCI_DEVICE(dev);
2173
2174 vms->iommu = VIRT_IOMMU_VIRTIO;
2175 vms->virtio_iommu_bdf = pci_get_bdf(pdev);
0fbddcec 2176 create_virtio_iommu_dt_bindings(vms);
70e89132 2177 }
1f283ae1
EA
2178}
2179
539533b8
SK
2180static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2181 DeviceState *dev, Error **errp)
2182{
2183 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2184 Error *local_err = NULL;
2185
2186 if (!vms->acpi_dev) {
2187 error_setg(&local_err,
2188 "memory hotplug is not enabled: missing acpi-ged device");
2189 goto out;
2190 }
2191
2192 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2193 error_setg(&local_err,
2194 "nvdimm device hot unplug is not supported yet.");
2195 goto out;
2196 }
2197
2198 hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2199 &local_err);
2200out:
2201 error_propagate(errp, local_err);
2202}
2203
2204static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2205 DeviceState *dev, Error **errp)
2206{
2207 VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2208 Error *local_err = NULL;
2209
2210 hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2211 if (local_err) {
2212 goto out;
2213 }
2214
2215 pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2216 qdev_unrealize(dev);
2217
2218out:
2219 error_propagate(errp, local_err);
2220}
2221
1f283ae1
EA
2222static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2223 DeviceState *dev, Error **errp)
2224{
539533b8
SK
2225 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2226 virt_dimm_unplug_request(hotplug_dev, dev, errp);
2227 } else {
2228 error_setg(errp, "device unplug request for unsupported device"
2229 " type: %s", object_get_typename(OBJECT(dev)));
2230 }
2231}
2232
2233static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2234 DeviceState *dev, Error **errp)
2235{
2236 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2237 virt_dimm_unplug(hotplug_dev, dev, errp);
2238 } else {
2239 error_setg(errp, "virt: device unplug for unsupported device"
2240 " type: %s", object_get_typename(OBJECT(dev)));
2241 }
a3fc8396
IM
2242}
2243
2244static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2245 DeviceState *dev)
2246{
1f283ae1
EA
2247 if (object_dynamic_cast(OBJECT(dev), TYPE_SYS_BUS_DEVICE) ||
2248 (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
a3fc8396
IM
2249 return HOTPLUG_HANDLER(machine);
2250 }
70e89132
EA
2251 if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2252 VirtMachineState *vms = VIRT_MACHINE(machine);
a3fc8396 2253
17e89077 2254 if (!vms->bootinfo.firmware_loaded || !virt_is_acpi_enabled(vms)) {
70e89132
EA
2255 return HOTPLUG_HANDLER(machine);
2256 }
2257 }
a3fc8396
IM
2258 return NULL;
2259}
2260
c9650222
EA
2261/*
2262 * for arm64 kvm_type [7-0] encodes the requested number of bits
2263 * in the IPA address space
2264 */
2265static int virt_kvm_type(MachineState *ms, const char *type_str)
2266{
2267 VirtMachineState *vms = VIRT_MACHINE(ms);
2268 int max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms);
2269 int requested_pa_size;
2270
2271 /* we freeze the memory map to compute the highest gpa */
2272 virt_set_memmap(vms);
2273
2274 requested_pa_size = 64 - clz64(vms->highest_gpa);
2275
2276 if (requested_pa_size > max_vm_pa_size) {
2277 error_report("-m and ,maxmem option values "
2278 "require an IPA range (%d bits) larger than "
2279 "the one supported by the host (%d bits)",
2280 requested_pa_size, max_vm_pa_size);
2281 exit(1);
2282 }
2283 /*
2284 * By default we return 0 which corresponds to an implicit legacy
2285 * 40b IPA setting. Otherwise we return the actual requested PA
2286 * logsize
2287 */
2288 return requested_pa_size > 40 ? requested_pa_size : 0;
2289}
2290
ed796373
WH
2291static void virt_machine_class_init(ObjectClass *oc, void *data)
2292{
9c94d8e6 2293 MachineClass *mc = MACHINE_CLASS(oc);
a3fc8396 2294 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
9c94d8e6
WH
2295
2296 mc->init = machvirt_init;
b10fbd53
EA
2297 /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2298 * The value may be reduced later when we have more information about the
9c94d8e6
WH
2299 * configuration of the particular instance.
2300 */
b10fbd53 2301 mc->max_cpus = 512;
6f2062b9
EH
2302 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2303 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
94692dcd 2304 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
4ebc0b61 2305 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
c294ac32 2306 machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
9c94d8e6
WH
2307 mc->block_default_type = IF_VIRTIO;
2308 mc->no_cdrom = 1;
2309 mc->pci_allow_0_address = true;
a2519ad1
PM
2310 /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2311 mc->minimum_page_bits = 12;
17d3d0e2 2312 mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
ea089eeb 2313 mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
ba1ba5cc 2314 mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
79e07936 2315 mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
c9650222 2316 mc->kvm_type = virt_kvm_type;
debbdc00 2317 assert(!mc->get_hotplug_handler);
a3fc8396 2318 mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
1f283ae1 2319 hc->pre_plug = virt_machine_device_pre_plug_cb;
a3fc8396 2320 hc->plug = virt_machine_device_plug_cb;
1f283ae1 2321 hc->unplug_request = virt_machine_device_unplug_request_cb;
539533b8 2322 hc->unplug = virt_machine_device_unplug_cb;
cd5ff833 2323 mc->numa_mem_supported = true;
c2505d1c 2324 mc->nvdimm_supported = true;
442da7dc 2325 mc->auto_enable_numa_with_memhp = true;
a72f6805 2326 mc->default_ram_id = "mach-virt.ram";
17e89077
GH
2327
2328 object_class_property_add(oc, "acpi", "OnOffAuto",
2329 virt_get_acpi, virt_set_acpi,
d2623129 2330 NULL, NULL);
17e89077 2331 object_class_property_set_description(oc, "acpi",
7eecec7d 2332 "Enable ACPI");
ed796373
WH
2333}
2334
95159760 2335static void virt_instance_init(Object *obj)
083a5890
GB
2336{
2337 VirtMachineState *vms = VIRT_MACHINE(obj);
ccc11b02 2338 VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
083a5890 2339
2d710006
PM
2340 /* EL3 is disabled by default on virt: this makes us consistent
2341 * between KVM and TCG for this board, and it also allows us to
2342 * boot UEFI blobs which assume no TrustZone support.
2343 */
2344 vms->secure = false;
083a5890 2345 object_property_add_bool(obj, "secure", virt_get_secure,
d2623129 2346 virt_set_secure);
083a5890
GB
2347 object_property_set_description(obj, "secure",
2348 "Set on/off to enable/disable the ARM "
7eecec7d 2349 "Security Extensions (TrustZone)");
5125f9cd 2350
f29cacfb
PM
2351 /* EL2 is also disabled by default, for similar reasons */
2352 vms->virt = false;
2353 object_property_add_bool(obj, "virtualization", virt_get_virt,
d2623129 2354 virt_set_virt);
f29cacfb
PM
2355 object_property_set_description(obj, "virtualization",
2356 "Set on/off to enable/disable emulating a "
2357 "guest CPU which implements the ARM "
7eecec7d 2358 "Virtualization Extensions");
f29cacfb 2359
5125f9cd
PF
2360 /* High memory is enabled by default */
2361 vms->highmem = true;
2362 object_property_add_bool(obj, "highmem", virt_get_highmem,
d2623129 2363 virt_set_highmem);
5125f9cd
PF
2364 object_property_set_description(obj, "highmem",
2365 "Set on/off to enable/disable using "
7eecec7d 2366 "physical address space above 32 bits");
36bf4ec8 2367 vms->gic_version = VIRT_GIC_VERSION_NOSEL;
b92ad394 2368 object_property_add_str(obj, "gic-version", virt_get_gic_version,
d2623129 2369 virt_set_gic_version);
b92ad394
PF
2370 object_property_set_description(obj, "gic-version",
2371 "Set GIC version. "
7eecec7d 2372 "Valid values are 2, 3, host and max");
9ac4ef77 2373
17ec075a
EA
2374 vms->highmem_ecam = !vmc->no_highmem_ecam;
2375
ccc11b02
EA
2376 if (vmc->no_its) {
2377 vms->its = false;
2378 } else {
2379 /* Default allows ITS instantiation */
2380 vms->its = true;
2381 object_property_add_bool(obj, "its", virt_get_its,
d2623129 2382 virt_set_its);
ccc11b02
EA
2383 object_property_set_description(obj, "its",
2384 "Set on/off to enable/disable "
7eecec7d 2385 "ITS instantiation");
ccc11b02
EA
2386 }
2387
e24e3454
EA
2388 /* Default disallows iommu instantiation */
2389 vms->iommu = VIRT_IOMMU_NONE;
d2623129 2390 object_property_add_str(obj, "iommu", virt_get_iommu, virt_set_iommu);
e24e3454
EA
2391 object_property_set_description(obj, "iommu",
2392 "Set the IOMMU type. "
7eecec7d 2393 "Valid values are none and smmuv3");
e24e3454 2394
2afa8c85
DG
2395 /* Default disallows RAS instantiation */
2396 vms->ras = false;
2397 object_property_add_bool(obj, "ras", virt_get_ras,
d2623129 2398 virt_set_ras);
2afa8c85
DG
2399 object_property_set_description(obj, "ras",
2400 "Set on/off to enable/disable reporting host memory errors "
7eecec7d 2401 "to a KVM guest using ACPI and guest external abort exceptions");
2afa8c85 2402
9ac4ef77 2403 vms->irqmap = a15irqmap;
e0561e60
MA
2404
2405 virt_flash_create(vms);
083a5890
GB
2406}
2407
95159760
EH
2408static const TypeInfo virt_machine_info = {
2409 .name = TYPE_VIRT_MACHINE,
2410 .parent = TYPE_MACHINE,
2411 .abstract = true,
2412 .instance_size = sizeof(VirtMachineState),
2413 .class_size = sizeof(VirtMachineClass),
2414 .class_init = virt_machine_class_init,
bbac02f1 2415 .instance_init = virt_instance_init,
95159760
EH
2416 .interfaces = (InterfaceInfo[]) {
2417 { TYPE_HOTPLUG_HANDLER },
2418 { }
2419 },
2420};
2421
2422static void machvirt_machine_init(void)
2423{
2424 type_register_static(&virt_machine_info);
2425}
2426type_init(machvirt_machine_init);
2427
541aaa1d
CH
2428static void virt_machine_5_1_options(MachineClass *mc)
2429{
2430}
2431DEFINE_VIRT_MACHINE_AS_LATEST(5, 1)
2432
3eb74d20
CH
2433static void virt_machine_5_0_options(MachineClass *mc)
2434{
541aaa1d 2435 virt_machine_5_1_options(mc);
c6228807 2436 compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3eb74d20 2437}
541aaa1d 2438DEFINE_VIRT_MACHINE(5, 0)
3eb74d20 2439
9aec2e52
CH
2440static void virt_machine_4_2_options(MachineClass *mc)
2441{
dea101a1
AJ
2442 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2443
fa7c8e92 2444 virt_machine_5_0_options(mc);
5f258577 2445 compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
dea101a1 2446 vmc->kvm_no_adjvtime = true;
9aec2e52 2447}
3eb74d20 2448DEFINE_VIRT_MACHINE(4, 2)
9aec2e52 2449
9bf2650b
CH
2450static void virt_machine_4_1_options(MachineClass *mc)
2451{
cff51ac9
SK
2452 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2453
9aec2e52
CH
2454 virt_machine_4_2_options(mc);
2455 compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
cff51ac9 2456 vmc->no_ged = true;
442da7dc 2457 mc->auto_enable_numa_with_memhp = false;
9bf2650b 2458}
9aec2e52 2459DEFINE_VIRT_MACHINE(4, 1)
9bf2650b 2460
84e060bf
AW
2461static void virt_machine_4_0_options(MachineClass *mc)
2462{
9bf2650b
CH
2463 virt_machine_4_1_options(mc);
2464 compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
84e060bf 2465}
9bf2650b 2466DEFINE_VIRT_MACHINE(4, 0)
84e060bf 2467
22907d2b
AJ
2468static void virt_machine_3_1_options(MachineClass *mc)
2469{
84e060bf 2470 virt_machine_4_0_options(mc);
abd93cc7 2471 compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
22907d2b 2472}
84e060bf 2473DEFINE_VIRT_MACHINE(3, 1)
22907d2b 2474
8ae9a1ca
EA
2475static void virt_machine_3_0_options(MachineClass *mc)
2476{
22907d2b 2477 virt_machine_3_1_options(mc);
ddb3235d 2478 compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
8ae9a1ca 2479}
22907d2b
AJ
2480DEFINE_VIRT_MACHINE(3, 0)
2481
a2a05159
PM
2482static void virt_machine_2_12_options(MachineClass *mc)
2483{
17ec075a
EA
2484 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2485
8ae9a1ca 2486 virt_machine_3_0_options(mc);
0d47310b 2487 compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
17ec075a 2488 vmc->no_highmem_ecam = true;
b10fbd53 2489 mc->max_cpus = 255;
a2a05159 2490}
8ae9a1ca 2491DEFINE_VIRT_MACHINE(2, 12)
a2a05159 2492
79283dda
EA
2493static void virt_machine_2_11_options(MachineClass *mc)
2494{
dfadc3bf
WH
2495 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2496
a2a05159 2497 virt_machine_2_12_options(mc);
43df70a9 2498 compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
dfadc3bf 2499 vmc->smbios_old_sys_ver = true;
79283dda 2500}
a2a05159 2501DEFINE_VIRT_MACHINE(2, 11)
79283dda 2502
f22ab6cb
EA
2503static void virt_machine_2_10_options(MachineClass *mc)
2504{
79283dda 2505 virt_machine_2_11_options(mc);
503224f4 2506 compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
846690de
PM
2507 /* before 2.11 we never faulted accesses to bad addresses */
2508 mc->ignore_memory_transaction_failures = true;
f22ab6cb 2509}
79283dda 2510DEFINE_VIRT_MACHINE(2, 10)
f22ab6cb 2511
e353aac5
PM
2512static void virt_machine_2_9_options(MachineClass *mc)
2513{
f22ab6cb 2514 virt_machine_2_10_options(mc);
3e803152 2515 compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
e353aac5 2516}
f22ab6cb 2517DEFINE_VIRT_MACHINE(2, 9)
e353aac5 2518
96b0439b
AJ
2519static void virt_machine_2_8_options(MachineClass *mc)
2520{
156bc9a5
PM
2521 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2522
e353aac5 2523 virt_machine_2_9_options(mc);
edc24ccd 2524 compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
156bc9a5
PM
2525 /* For 2.8 and earlier we falsely claimed in the DT that
2526 * our timers were edge-triggered, not level-triggered.
2527 */
2528 vmc->claim_edge_triggered_timers = true;
96b0439b 2529}
e353aac5 2530DEFINE_VIRT_MACHINE(2, 8)
96b0439b 2531
1287f2b3
AJ
2532static void virt_machine_2_7_options(MachineClass *mc)
2533{
2231f69b
AJ
2534 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2535
96b0439b 2536 virt_machine_2_8_options(mc);
5a995064 2537 compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2231f69b
AJ
2538 /* ITS was introduced with 2.8 */
2539 vmc->no_its = true;
a2519ad1
PM
2540 /* Stick with 1K pages for migration compatibility */
2541 mc->minimum_page_bits = 0;
1287f2b3 2542}
96b0439b 2543DEFINE_VIRT_MACHINE(2, 7)
1287f2b3 2544
ab093c3c 2545static void virt_machine_2_6_options(MachineClass *mc)
c2919690 2546{
95eb49c8
AJ
2547 VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2548
1287f2b3 2549 virt_machine_2_7_options(mc);
ff8f261f 2550 compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
95eb49c8 2551 vmc->disallow_affinity_adjustment = true;
1141d1eb
WH
2552 /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2553 vmc->no_pmu = true;
c2919690 2554}
1287f2b3 2555DEFINE_VIRT_MACHINE(2, 6)