xref: /openbmc/qemu/hw/ppc/pnv.c (revision 4b9fa0b4)
1 /*
2  * QEMU PowerPC PowerNV machine model
3  *
4  * Copyright (c) 2016, IBM Corporation.
5  *
6  * This library is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2 of the License, or (at your option) any later version.
10  *
11  * This library is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
18  */
19 
20 #include "qemu/osdep.h"
21 #include "qemu-common.h"
22 #include "qemu/units.h"
23 #include "qapi/error.h"
24 #include "sysemu/sysemu.h"
25 #include "sysemu/numa.h"
26 #include "sysemu/reset.h"
27 #include "sysemu/runstate.h"
28 #include "sysemu/cpus.h"
29 #include "sysemu/device_tree.h"
30 #include "target/ppc/cpu.h"
31 #include "qemu/log.h"
32 #include "hw/ppc/fdt.h"
33 #include "hw/ppc/ppc.h"
34 #include "hw/ppc/pnv.h"
35 #include "hw/ppc/pnv_core.h"
36 #include "hw/loader.h"
37 #include "exec/address-spaces.h"
38 #include "qapi/visitor.h"
39 #include "monitor/monitor.h"
40 #include "hw/intc/intc.h"
41 #include "hw/ipmi/ipmi.h"
42 #include "target/ppc/mmu-hash64.h"
43 
44 #include "hw/ppc/xics.h"
45 #include "hw/qdev-properties.h"
46 #include "hw/ppc/pnv_xscom.h"
47 
48 #include "hw/isa/isa.h"
49 #include "hw/boards.h"
50 #include "hw/char/serial.h"
51 #include "hw/rtc/mc146818rtc.h"
52 
53 #include <libfdt.h>
54 
55 #define FDT_MAX_SIZE            (1 * MiB)
56 
57 #define FW_FILE_NAME            "skiboot.lid"
58 #define FW_LOAD_ADDR            0x0
59 #define FW_MAX_SIZE             (4 * MiB)
60 
61 #define KERNEL_LOAD_ADDR        0x20000000
62 #define KERNEL_MAX_SIZE         (256 * MiB)
63 #define INITRD_LOAD_ADDR        0x60000000
64 #define INITRD_MAX_SIZE         (256 * MiB)
65 
66 static const char *pnv_chip_core_typename(const PnvChip *o)
67 {
68     const char *chip_type = object_class_get_name(object_get_class(OBJECT(o)));
69     int len = strlen(chip_type) - strlen(PNV_CHIP_TYPE_SUFFIX);
70     char *s = g_strdup_printf(PNV_CORE_TYPE_NAME("%.*s"), len, chip_type);
71     const char *core_type = object_class_get_name(object_class_by_name(s));
72     g_free(s);
73     return core_type;
74 }
75 
76 /*
77  * On Power Systems E880 (POWER8), the max cpus (threads) should be :
78  *     4 * 4 sockets * 12 cores * 8 threads = 1536
79  * Let's make it 2^11
80  */
81 #define MAX_CPUS                2048
82 
83 /*
84  * Memory nodes are created by hostboot, one for each range of memory
85  * that has a different "affinity". In practice, it means one range
86  * per chip.
87  */
88 static void pnv_dt_memory(void *fdt, int chip_id, hwaddr start, hwaddr size)
89 {
90     char *mem_name;
91     uint64_t mem_reg_property[2];
92     int off;
93 
94     mem_reg_property[0] = cpu_to_be64(start);
95     mem_reg_property[1] = cpu_to_be64(size);
96 
97     mem_name = g_strdup_printf("memory@%"HWADDR_PRIx, start);
98     off = fdt_add_subnode(fdt, 0, mem_name);
99     g_free(mem_name);
100 
101     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
102     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
103                        sizeof(mem_reg_property))));
104     _FDT((fdt_setprop_cell(fdt, off, "ibm,chip-id", chip_id)));
105 }
106 
107 static int get_cpus_node(void *fdt)
108 {
109     int cpus_offset = fdt_path_offset(fdt, "/cpus");
110 
111     if (cpus_offset < 0) {
112         cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
113         if (cpus_offset) {
114             _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
115             _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
116         }
117     }
118     _FDT(cpus_offset);
119     return cpus_offset;
120 }
121 
122 /*
123  * The PowerNV cores (and threads) need to use real HW ids and not an
124  * incremental index like it has been done on other platforms. This HW
125  * id is stored in the CPU PIR, it is used to create cpu nodes in the
126  * device tree, used in XSCOM to address cores and in interrupt
127  * servers.
128  */
129 static void pnv_dt_core(PnvChip *chip, PnvCore *pc, void *fdt)
130 {
131     PowerPCCPU *cpu = pc->threads[0];
132     CPUState *cs = CPU(cpu);
133     DeviceClass *dc = DEVICE_GET_CLASS(cs);
134     int smt_threads = CPU_CORE(pc)->nr_threads;
135     CPUPPCState *env = &cpu->env;
136     PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
137     uint32_t servers_prop[smt_threads];
138     int i;
139     uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
140                        0xffffffff, 0xffffffff};
141     uint32_t tbfreq = PNV_TIMEBASE_FREQ;
142     uint32_t cpufreq = 1000000000;
143     uint32_t page_sizes_prop[64];
144     size_t page_sizes_prop_size;
145     const uint8_t pa_features[] = { 24, 0,
146                                     0xf6, 0x3f, 0xc7, 0xc0, 0x80, 0xf0,
147                                     0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
148                                     0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
149                                     0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
150     int offset;
151     char *nodename;
152     int cpus_offset = get_cpus_node(fdt);
153 
154     nodename = g_strdup_printf("%s@%x", dc->fw_name, pc->pir);
155     offset = fdt_add_subnode(fdt, cpus_offset, nodename);
156     _FDT(offset);
157     g_free(nodename);
158 
159     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", chip->chip_id)));
160 
161     _FDT((fdt_setprop_cell(fdt, offset, "reg", pc->pir)));
162     _FDT((fdt_setprop_cell(fdt, offset, "ibm,pir", pc->pir)));
163     _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
164 
165     _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
166     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
167                             env->dcache_line_size)));
168     _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
169                             env->dcache_line_size)));
170     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
171                             env->icache_line_size)));
172     _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
173                             env->icache_line_size)));
174 
175     if (pcc->l1_dcache_size) {
176         _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
177                                pcc->l1_dcache_size)));
178     } else {
179         warn_report("Unknown L1 dcache size for cpu");
180     }
181     if (pcc->l1_icache_size) {
182         _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
183                                pcc->l1_icache_size)));
184     } else {
185         warn_report("Unknown L1 icache size for cpu");
186     }
187 
188     _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
189     _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
190     _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size",
191                            cpu->hash64_opts->slb_size)));
192     _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
193     _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
194 
195     if (env->spr_cb[SPR_PURR].oea_read) {
196         _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
197     }
198 
199     if (ppc_hash64_has(cpu, PPC_HASH64_1TSEG)) {
200         _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
201                            segs, sizeof(segs))));
202     }
203 
204     /*
205      * Advertise VMX/VSX (vector extensions) if available
206      *   0 / no property == no vector extensions
207      *   1               == VMX / Altivec available
208      *   2               == VSX available
209      */
210     if (env->insns_flags & PPC_ALTIVEC) {
211         uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
212 
213         _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
214     }
215 
216     /*
217      * Advertise DFP (Decimal Floating Point) if available
218      *   0 / no property == no DFP
219      *   1               == DFP available
220      */
221     if (env->insns_flags2 & PPC2_DFP) {
222         _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
223     }
224 
225     page_sizes_prop_size = ppc_create_page_sizes_prop(cpu, page_sizes_prop,
226                                                       sizeof(page_sizes_prop));
227     if (page_sizes_prop_size) {
228         _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
229                            page_sizes_prop, page_sizes_prop_size)));
230     }
231 
232     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features",
233                        pa_features, sizeof(pa_features))));
234 
235     /* Build interrupt servers properties */
236     for (i = 0; i < smt_threads; i++) {
237         servers_prop[i] = cpu_to_be32(pc->pir + i);
238     }
239     _FDT((fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
240                        servers_prop, sizeof(servers_prop))));
241 }
242 
243 static void pnv_dt_icp(PnvChip *chip, void *fdt, uint32_t pir,
244                        uint32_t nr_threads)
245 {
246     uint64_t addr = PNV_ICP_BASE(chip) | (pir << 12);
247     char *name;
248     const char compat[] = "IBM,power8-icp\0IBM,ppc-xicp";
249     uint32_t irange[2], i, rsize;
250     uint64_t *reg;
251     int offset;
252 
253     irange[0] = cpu_to_be32(pir);
254     irange[1] = cpu_to_be32(nr_threads);
255 
256     rsize = sizeof(uint64_t) * 2 * nr_threads;
257     reg = g_malloc(rsize);
258     for (i = 0; i < nr_threads; i++) {
259         reg[i * 2] = cpu_to_be64(addr | ((pir + i) * 0x1000));
260         reg[i * 2 + 1] = cpu_to_be64(0x1000);
261     }
262 
263     name = g_strdup_printf("interrupt-controller@%"PRIX64, addr);
264     offset = fdt_add_subnode(fdt, 0, name);
265     _FDT(offset);
266     g_free(name);
267 
268     _FDT((fdt_setprop(fdt, offset, "compatible", compat, sizeof(compat))));
269     _FDT((fdt_setprop(fdt, offset, "reg", reg, rsize)));
270     _FDT((fdt_setprop_string(fdt, offset, "device_type",
271                               "PowerPC-External-Interrupt-Presentation")));
272     _FDT((fdt_setprop(fdt, offset, "interrupt-controller", NULL, 0)));
273     _FDT((fdt_setprop(fdt, offset, "ibm,interrupt-server-ranges",
274                        irange, sizeof(irange))));
275     _FDT((fdt_setprop_cell(fdt, offset, "#interrupt-cells", 1)));
276     _FDT((fdt_setprop_cell(fdt, offset, "#address-cells", 0)));
277     g_free(reg);
278 }
279 
280 static void pnv_chip_power8_dt_populate(PnvChip *chip, void *fdt)
281 {
282     const char *typename = pnv_chip_core_typename(chip);
283     size_t typesize = object_type_get_instance_size(typename);
284     int i;
285 
286     pnv_dt_xscom(chip, fdt, 0);
287 
288     for (i = 0; i < chip->nr_cores; i++) {
289         PnvCore *pnv_core = PNV_CORE(chip->cores + i * typesize);
290 
291         pnv_dt_core(chip, pnv_core, fdt);
292 
293         /* Interrupt Control Presenters (ICP). One per core. */
294         pnv_dt_icp(chip, fdt, pnv_core->pir, CPU_CORE(pnv_core)->nr_threads);
295     }
296 
297     if (chip->ram_size) {
298         pnv_dt_memory(fdt, chip->chip_id, chip->ram_start, chip->ram_size);
299     }
300 }
301 
302 static void pnv_chip_power9_dt_populate(PnvChip *chip, void *fdt)
303 {
304     const char *typename = pnv_chip_core_typename(chip);
305     size_t typesize = object_type_get_instance_size(typename);
306     int i;
307 
308     pnv_dt_xscom(chip, fdt, 0);
309 
310     for (i = 0; i < chip->nr_cores; i++) {
311         PnvCore *pnv_core = PNV_CORE(chip->cores + i * typesize);
312 
313         pnv_dt_core(chip, pnv_core, fdt);
314     }
315 
316     if (chip->ram_size) {
317         pnv_dt_memory(fdt, chip->chip_id, chip->ram_start, chip->ram_size);
318     }
319 
320     pnv_dt_lpc(chip, fdt, 0);
321 }
322 
323 static void pnv_dt_rtc(ISADevice *d, void *fdt, int lpc_off)
324 {
325     uint32_t io_base = d->ioport_id;
326     uint32_t io_regs[] = {
327         cpu_to_be32(1),
328         cpu_to_be32(io_base),
329         cpu_to_be32(2)
330     };
331     char *name;
332     int node;
333 
334     name = g_strdup_printf("%s@i%x", qdev_fw_name(DEVICE(d)), io_base);
335     node = fdt_add_subnode(fdt, lpc_off, name);
336     _FDT(node);
337     g_free(name);
338 
339     _FDT((fdt_setprop(fdt, node, "reg", io_regs, sizeof(io_regs))));
340     _FDT((fdt_setprop_string(fdt, node, "compatible", "pnpPNP,b00")));
341 }
342 
343 static void pnv_dt_serial(ISADevice *d, void *fdt, int lpc_off)
344 {
345     const char compatible[] = "ns16550\0pnpPNP,501";
346     uint32_t io_base = d->ioport_id;
347     uint32_t io_regs[] = {
348         cpu_to_be32(1),
349         cpu_to_be32(io_base),
350         cpu_to_be32(8)
351     };
352     char *name;
353     int node;
354 
355     name = g_strdup_printf("%s@i%x", qdev_fw_name(DEVICE(d)), io_base);
356     node = fdt_add_subnode(fdt, lpc_off, name);
357     _FDT(node);
358     g_free(name);
359 
360     _FDT((fdt_setprop(fdt, node, "reg", io_regs, sizeof(io_regs))));
361     _FDT((fdt_setprop(fdt, node, "compatible", compatible,
362                       sizeof(compatible))));
363 
364     _FDT((fdt_setprop_cell(fdt, node, "clock-frequency", 1843200)));
365     _FDT((fdt_setprop_cell(fdt, node, "current-speed", 115200)));
366     _FDT((fdt_setprop_cell(fdt, node, "interrupts", d->isairq[0])));
367     _FDT((fdt_setprop_cell(fdt, node, "interrupt-parent",
368                            fdt_get_phandle(fdt, lpc_off))));
369 
370     /* This is needed by Linux */
371     _FDT((fdt_setprop_string(fdt, node, "device_type", "serial")));
372 }
373 
374 static void pnv_dt_ipmi_bt(ISADevice *d, void *fdt, int lpc_off)
375 {
376     const char compatible[] = "bt\0ipmi-bt";
377     uint32_t io_base;
378     uint32_t io_regs[] = {
379         cpu_to_be32(1),
380         0, /* 'io_base' retrieved from the 'ioport' property of 'isa-ipmi-bt' */
381         cpu_to_be32(3)
382     };
383     uint32_t irq;
384     char *name;
385     int node;
386 
387     io_base = object_property_get_int(OBJECT(d), "ioport", &error_fatal);
388     io_regs[1] = cpu_to_be32(io_base);
389 
390     irq = object_property_get_int(OBJECT(d), "irq", &error_fatal);
391 
392     name = g_strdup_printf("%s@i%x", qdev_fw_name(DEVICE(d)), io_base);
393     node = fdt_add_subnode(fdt, lpc_off, name);
394     _FDT(node);
395     g_free(name);
396 
397     _FDT((fdt_setprop(fdt, node, "reg", io_regs, sizeof(io_regs))));
398     _FDT((fdt_setprop(fdt, node, "compatible", compatible,
399                       sizeof(compatible))));
400 
401     /* Mark it as reserved to avoid Linux trying to claim it */
402     _FDT((fdt_setprop_string(fdt, node, "status", "reserved")));
403     _FDT((fdt_setprop_cell(fdt, node, "interrupts", irq)));
404     _FDT((fdt_setprop_cell(fdt, node, "interrupt-parent",
405                            fdt_get_phandle(fdt, lpc_off))));
406 }
407 
408 typedef struct ForeachPopulateArgs {
409     void *fdt;
410     int offset;
411 } ForeachPopulateArgs;
412 
413 static int pnv_dt_isa_device(DeviceState *dev, void *opaque)
414 {
415     ForeachPopulateArgs *args = opaque;
416     ISADevice *d = ISA_DEVICE(dev);
417 
418     if (object_dynamic_cast(OBJECT(dev), TYPE_MC146818_RTC)) {
419         pnv_dt_rtc(d, args->fdt, args->offset);
420     } else if (object_dynamic_cast(OBJECT(dev), TYPE_ISA_SERIAL)) {
421         pnv_dt_serial(d, args->fdt, args->offset);
422     } else if (object_dynamic_cast(OBJECT(dev), "isa-ipmi-bt")) {
423         pnv_dt_ipmi_bt(d, args->fdt, args->offset);
424     } else {
425         error_report("unknown isa device %s@i%x", qdev_fw_name(dev),
426                      d->ioport_id);
427     }
428 
429     return 0;
430 }
431 
432 /*
433  * The default LPC bus of a multichip system is on chip 0. It's
434  * recognized by the firmware (skiboot) using a "primary" property.
435  */
436 static void pnv_dt_isa(PnvMachineState *pnv, void *fdt)
437 {
438     int isa_offset = fdt_path_offset(fdt, pnv->chips[0]->dt_isa_nodename);
439     ForeachPopulateArgs args = {
440         .fdt = fdt,
441         .offset = isa_offset,
442     };
443     uint32_t phandle;
444 
445     _FDT((fdt_setprop(fdt, isa_offset, "primary", NULL, 0)));
446 
447     phandle = qemu_fdt_alloc_phandle(fdt);
448     assert(phandle > 0);
449     _FDT((fdt_setprop_cell(fdt, isa_offset, "phandle", phandle)));
450 
451     /*
452      * ISA devices are not necessarily parented to the ISA bus so we
453      * can not use object_child_foreach()
454      */
455     qbus_walk_children(BUS(pnv->isa_bus), pnv_dt_isa_device, NULL, NULL, NULL,
456                        &args);
457 }
458 
459 static void pnv_dt_power_mgt(void *fdt)
460 {
461     int off;
462 
463     off = fdt_add_subnode(fdt, 0, "ibm,opal");
464     off = fdt_add_subnode(fdt, off, "power-mgt");
465 
466     _FDT(fdt_setprop_cell(fdt, off, "ibm,enabled-stop-levels", 0xc0000000));
467 }
468 
469 static void *pnv_dt_create(MachineState *machine)
470 {
471     const char plat_compat8[] = "qemu,powernv8\0qemu,powernv\0ibm,powernv";
472     const char plat_compat9[] = "qemu,powernv9\0ibm,powernv";
473     PnvMachineState *pnv = PNV_MACHINE(machine);
474     void *fdt;
475     char *buf;
476     int off;
477     int i;
478 
479     fdt = g_malloc0(FDT_MAX_SIZE);
480     _FDT((fdt_create_empty_tree(fdt, FDT_MAX_SIZE)));
481 
482     /* Root node */
483     _FDT((fdt_setprop_cell(fdt, 0, "#address-cells", 0x2)));
484     _FDT((fdt_setprop_cell(fdt, 0, "#size-cells", 0x2)));
485     _FDT((fdt_setprop_string(fdt, 0, "model",
486                              "IBM PowerNV (emulated by qemu)")));
487     if (pnv_is_power9(pnv)) {
488         _FDT((fdt_setprop(fdt, 0, "compatible", plat_compat9,
489                           sizeof(plat_compat9))));
490     } else {
491         _FDT((fdt_setprop(fdt, 0, "compatible", plat_compat8,
492                           sizeof(plat_compat8))));
493     }
494 
495 
496     buf =  qemu_uuid_unparse_strdup(&qemu_uuid);
497     _FDT((fdt_setprop_string(fdt, 0, "vm,uuid", buf)));
498     if (qemu_uuid_set) {
499         _FDT((fdt_property_string(fdt, "system-id", buf)));
500     }
501     g_free(buf);
502 
503     off = fdt_add_subnode(fdt, 0, "chosen");
504     if (machine->kernel_cmdline) {
505         _FDT((fdt_setprop_string(fdt, off, "bootargs",
506                                  machine->kernel_cmdline)));
507     }
508 
509     if (pnv->initrd_size) {
510         uint32_t start_prop = cpu_to_be32(pnv->initrd_base);
511         uint32_t end_prop = cpu_to_be32(pnv->initrd_base + pnv->initrd_size);
512 
513         _FDT((fdt_setprop(fdt, off, "linux,initrd-start",
514                                &start_prop, sizeof(start_prop))));
515         _FDT((fdt_setprop(fdt, off, "linux,initrd-end",
516                                &end_prop, sizeof(end_prop))));
517     }
518 
519     /* Populate device tree for each chip */
520     for (i = 0; i < pnv->num_chips; i++) {
521         PNV_CHIP_GET_CLASS(pnv->chips[i])->dt_populate(pnv->chips[i], fdt);
522     }
523 
524     /* Populate ISA devices on chip 0 */
525     pnv_dt_isa(pnv, fdt);
526 
527     if (pnv->bmc) {
528         pnv_dt_bmc_sensors(pnv->bmc, fdt);
529     }
530 
531     /* Create an extra node for power management on Power9 */
532     if (pnv_is_power9(pnv)) {
533         pnv_dt_power_mgt(fdt);
534     }
535 
536     return fdt;
537 }
538 
539 static void pnv_powerdown_notify(Notifier *n, void *opaque)
540 {
541     PnvMachineState *pnv = PNV_MACHINE(qdev_get_machine());
542 
543     if (pnv->bmc) {
544         pnv_bmc_powerdown(pnv->bmc);
545     }
546 }
547 
548 static void pnv_reset(MachineState *machine)
549 {
550     PnvMachineState *pnv = PNV_MACHINE(machine);
551     void *fdt;
552     Object *obj;
553 
554     qemu_devices_reset();
555 
556     /*
557      * OpenPOWER systems have a BMC, which can be defined on the
558      * command line with:
559      *
560      *   -device ipmi-bmc-sim,id=bmc0
561      *
562      * This is the internal simulator but it could also be an external
563      * BMC.
564      */
565     obj = object_resolve_path_type("", "ipmi-bmc-sim", NULL);
566     if (obj) {
567         pnv->bmc = IPMI_BMC(obj);
568     }
569 
570     fdt = pnv_dt_create(machine);
571 
572     /* Pack resulting tree */
573     _FDT((fdt_pack(fdt)));
574 
575     qemu_fdt_dumpdtb(fdt, fdt_totalsize(fdt));
576     cpu_physical_memory_write(PNV_FDT_ADDR, fdt, fdt_totalsize(fdt));
577 }
578 
579 static ISABus *pnv_chip_power8_isa_create(PnvChip *chip, Error **errp)
580 {
581     Pnv8Chip *chip8 = PNV8_CHIP(chip);
582     return pnv_lpc_isa_create(&chip8->lpc, true, errp);
583 }
584 
585 static ISABus *pnv_chip_power8nvl_isa_create(PnvChip *chip, Error **errp)
586 {
587     Pnv8Chip *chip8 = PNV8_CHIP(chip);
588     return pnv_lpc_isa_create(&chip8->lpc, false, errp);
589 }
590 
591 static ISABus *pnv_chip_power9_isa_create(PnvChip *chip, Error **errp)
592 {
593     Pnv9Chip *chip9 = PNV9_CHIP(chip);
594     return pnv_lpc_isa_create(&chip9->lpc, false, errp);
595 }
596 
597 static ISABus *pnv_isa_create(PnvChip *chip, Error **errp)
598 {
599     return PNV_CHIP_GET_CLASS(chip)->isa_create(chip, errp);
600 }
601 
602 static void pnv_chip_power8_pic_print_info(PnvChip *chip, Monitor *mon)
603 {
604     Pnv8Chip *chip8 = PNV8_CHIP(chip);
605 
606     ics_pic_print_info(&chip8->psi.ics, mon);
607 }
608 
609 static void pnv_chip_power9_pic_print_info(PnvChip *chip, Monitor *mon)
610 {
611     Pnv9Chip *chip9 = PNV9_CHIP(chip);
612 
613     pnv_xive_pic_print_info(&chip9->xive, mon);
614     pnv_psi_pic_print_info(&chip9->psi, mon);
615 }
616 
617 static bool pnv_match_cpu(const char *default_type, const char *cpu_type)
618 {
619     PowerPCCPUClass *ppc_default =
620         POWERPC_CPU_CLASS(object_class_by_name(default_type));
621     PowerPCCPUClass *ppc =
622         POWERPC_CPU_CLASS(object_class_by_name(cpu_type));
623 
624     return ppc_default->pvr_match(ppc_default, ppc->pvr);
625 }
626 
627 static void pnv_init(MachineState *machine)
628 {
629     PnvMachineState *pnv = PNV_MACHINE(machine);
630     MachineClass *mc = MACHINE_GET_CLASS(machine);
631     MemoryRegion *ram;
632     char *fw_filename;
633     long fw_size;
634     int i;
635     char *chip_typename;
636 
637     /* allocate RAM */
638     if (machine->ram_size < (1 * GiB)) {
639         warn_report("skiboot may not work with < 1GB of RAM");
640     }
641 
642     ram = g_new(MemoryRegion, 1);
643     memory_region_allocate_system_memory(ram, NULL, "pnv.ram",
644                                          machine->ram_size);
645     memory_region_add_subregion(get_system_memory(), 0, ram);
646 
647     /* load skiboot firmware  */
648     if (bios_name == NULL) {
649         bios_name = FW_FILE_NAME;
650     }
651 
652     fw_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
653     if (!fw_filename) {
654         error_report("Could not find OPAL firmware '%s'", bios_name);
655         exit(1);
656     }
657 
658     fw_size = load_image_targphys(fw_filename, FW_LOAD_ADDR, FW_MAX_SIZE);
659     if (fw_size < 0) {
660         error_report("Could not load OPAL firmware '%s'", fw_filename);
661         exit(1);
662     }
663     g_free(fw_filename);
664 
665     /* load kernel */
666     if (machine->kernel_filename) {
667         long kernel_size;
668 
669         kernel_size = load_image_targphys(machine->kernel_filename,
670                                           KERNEL_LOAD_ADDR, KERNEL_MAX_SIZE);
671         if (kernel_size < 0) {
672             error_report("Could not load kernel '%s'",
673                          machine->kernel_filename);
674             exit(1);
675         }
676     }
677 
678     /* load initrd */
679     if (machine->initrd_filename) {
680         pnv->initrd_base = INITRD_LOAD_ADDR;
681         pnv->initrd_size = load_image_targphys(machine->initrd_filename,
682                                   pnv->initrd_base, INITRD_MAX_SIZE);
683         if (pnv->initrd_size < 0) {
684             error_report("Could not load initial ram disk '%s'",
685                          machine->initrd_filename);
686             exit(1);
687         }
688     }
689 
690     /*
691      * Check compatibility of the specified CPU with the machine
692      * default.
693      */
694     if (!pnv_match_cpu(mc->default_cpu_type, machine->cpu_type)) {
695         error_report("invalid CPU model '%s' for %s machine",
696                      machine->cpu_type, mc->name);
697         exit(1);
698     }
699 
700     /* Create the processor chips */
701     i = strlen(machine->cpu_type) - strlen(POWERPC_CPU_TYPE_SUFFIX);
702     chip_typename = g_strdup_printf(PNV_CHIP_TYPE_NAME("%.*s"),
703                                     i, machine->cpu_type);
704     if (!object_class_by_name(chip_typename)) {
705         error_report("invalid chip model '%.*s' for %s machine",
706                      i, machine->cpu_type, mc->name);
707         exit(1);
708     }
709 
710     pnv->chips = g_new0(PnvChip *, pnv->num_chips);
711     for (i = 0; i < pnv->num_chips; i++) {
712         char chip_name[32];
713         Object *chip = object_new(chip_typename);
714 
715         pnv->chips[i] = PNV_CHIP(chip);
716 
717         /*
718          * TODO: put all the memory in one node on chip 0 until we find a
719          * way to specify different ranges for each chip
720          */
721         if (i == 0) {
722             object_property_set_int(chip, machine->ram_size, "ram-size",
723                                     &error_fatal);
724         }
725 
726         snprintf(chip_name, sizeof(chip_name), "chip[%d]", PNV_CHIP_HWID(i));
727         object_property_add_child(OBJECT(pnv), chip_name, chip, &error_fatal);
728         object_property_set_int(chip, PNV_CHIP_HWID(i), "chip-id",
729                                 &error_fatal);
730         object_property_set_int(chip, machine->smp.cores,
731                                 "nr-cores", &error_fatal);
732         object_property_set_bool(chip, true, "realized", &error_fatal);
733     }
734     g_free(chip_typename);
735 
736     /* Instantiate ISA bus on chip 0 */
737     pnv->isa_bus = pnv_isa_create(pnv->chips[0], &error_fatal);
738 
739     /* Create serial port */
740     serial_hds_isa_init(pnv->isa_bus, 0, MAX_ISA_SERIAL_PORTS);
741 
742     /* Create an RTC ISA device too */
743     mc146818_rtc_init(pnv->isa_bus, 2000, NULL);
744 
745     /*
746      * OpenPOWER systems use a IPMI SEL Event message to notify the
747      * host to powerdown
748      */
749     pnv->powerdown_notifier.notify = pnv_powerdown_notify;
750     qemu_register_powerdown_notifier(&pnv->powerdown_notifier);
751 }
752 
753 /*
754  *    0:21  Reserved - Read as zeros
755  *   22:24  Chip ID
756  *   25:28  Core number
757  *   29:31  Thread ID
758  */
759 static uint32_t pnv_chip_core_pir_p8(PnvChip *chip, uint32_t core_id)
760 {
761     return (chip->chip_id << 7) | (core_id << 3);
762 }
763 
764 static void pnv_chip_power8_intc_create(PnvChip *chip, PowerPCCPU *cpu,
765                                         Error **errp)
766 {
767     Error *local_err = NULL;
768     Object *obj;
769     PnvCPUState *pnv_cpu = pnv_cpu_state(cpu);
770 
771     obj = icp_create(OBJECT(cpu), TYPE_PNV_ICP, XICS_FABRIC(qdev_get_machine()),
772                      &local_err);
773     if (local_err) {
774         error_propagate(errp, local_err);
775         return;
776     }
777 
778     pnv_cpu->intc = obj;
779 }
780 
781 static void pnv_chip_power8_intc_reset(PnvChip *chip, PowerPCCPU *cpu)
782 {
783     PnvCPUState *pnv_cpu = pnv_cpu_state(cpu);
784 
785     icp_reset(ICP(pnv_cpu->intc));
786 }
787 
788 /*
789  *    0:48  Reserved - Read as zeroes
790  *   49:52  Node ID
791  *   53:55  Chip ID
792  *   56     Reserved - Read as zero
793  *   57:61  Core number
794  *   62:63  Thread ID
795  *
796  * We only care about the lower bits. uint32_t is fine for the moment.
797  */
798 static uint32_t pnv_chip_core_pir_p9(PnvChip *chip, uint32_t core_id)
799 {
800     return (chip->chip_id << 8) | (core_id << 2);
801 }
802 
803 static void pnv_chip_power9_intc_create(PnvChip *chip, PowerPCCPU *cpu,
804                                         Error **errp)
805 {
806     Pnv9Chip *chip9 = PNV9_CHIP(chip);
807     Error *local_err = NULL;
808     Object *obj;
809     PnvCPUState *pnv_cpu = pnv_cpu_state(cpu);
810 
811     /*
812      * The core creates its interrupt presenter but the XIVE interrupt
813      * controller object is initialized afterwards. Hopefully, it's
814      * only used at runtime.
815      */
816     obj = xive_tctx_create(OBJECT(cpu), XIVE_ROUTER(&chip9->xive), &local_err);
817     if (local_err) {
818         error_propagate(errp, local_err);
819         return;
820     }
821 
822     pnv_cpu->intc = obj;
823 }
824 
825 static void pnv_chip_power9_intc_reset(PnvChip *chip, PowerPCCPU *cpu)
826 {
827     PnvCPUState *pnv_cpu = pnv_cpu_state(cpu);
828 
829     xive_tctx_reset(XIVE_TCTX(pnv_cpu->intc));
830 }
831 
832 /*
833  * Allowed core identifiers on a POWER8 Processor Chip :
834  *
835  * <EX0 reserved>
836  *  EX1  - Venice only
837  *  EX2  - Venice only
838  *  EX3  - Venice only
839  *  EX4
840  *  EX5
841  *  EX6
842  * <EX7,8 reserved> <reserved>
843  *  EX9  - Venice only
844  *  EX10 - Venice only
845  *  EX11 - Venice only
846  *  EX12
847  *  EX13
848  *  EX14
849  * <EX15 reserved>
850  */
851 #define POWER8E_CORE_MASK  (0x7070ull)
852 #define POWER8_CORE_MASK   (0x7e7eull)
853 
854 /*
855  * POWER9 has 24 cores, ids starting at 0x0
856  */
857 #define POWER9_CORE_MASK   (0xffffffffffffffull)
858 
859 static void pnv_chip_power8_instance_init(Object *obj)
860 {
861     Pnv8Chip *chip8 = PNV8_CHIP(obj);
862 
863     object_initialize_child(obj, "psi",  &chip8->psi, sizeof(chip8->psi),
864                             TYPE_PNV8_PSI, &error_abort, NULL);
865     object_property_add_const_link(OBJECT(&chip8->psi), "xics",
866                                    OBJECT(qdev_get_machine()), &error_abort);
867 
868     object_initialize_child(obj, "lpc",  &chip8->lpc, sizeof(chip8->lpc),
869                             TYPE_PNV8_LPC, &error_abort, NULL);
870     object_property_add_const_link(OBJECT(&chip8->lpc), "psi",
871                                    OBJECT(&chip8->psi), &error_abort);
872 
873     object_initialize_child(obj, "occ",  &chip8->occ, sizeof(chip8->occ),
874                             TYPE_PNV8_OCC, &error_abort, NULL);
875     object_property_add_const_link(OBJECT(&chip8->occ), "psi",
876                                    OBJECT(&chip8->psi), &error_abort);
877 
878     object_initialize_child(obj, "homer",  &chip8->homer, sizeof(chip8->homer),
879                             TYPE_PNV8_HOMER, &error_abort, NULL);
880     object_property_add_const_link(OBJECT(&chip8->homer), "chip", obj,
881                                    &error_abort);
882 }
883 
884 static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
885  {
886     PnvChip *chip = PNV_CHIP(chip8);
887     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
888     const char *typename = pnv_chip_core_typename(chip);
889     size_t typesize = object_type_get_instance_size(typename);
890     int i, j;
891     char *name;
892     XICSFabric *xi = XICS_FABRIC(qdev_get_machine());
893 
894     name = g_strdup_printf("icp-%x", chip->chip_id);
895     memory_region_init(&chip8->icp_mmio, OBJECT(chip), name, PNV_ICP_SIZE);
896     sysbus_init_mmio(SYS_BUS_DEVICE(chip), &chip8->icp_mmio);
897     g_free(name);
898 
899     sysbus_mmio_map(SYS_BUS_DEVICE(chip), 1, PNV_ICP_BASE(chip));
900 
901     /* Map the ICP registers for each thread */
902     for (i = 0; i < chip->nr_cores; i++) {
903         PnvCore *pnv_core = PNV_CORE(chip->cores + i * typesize);
904         int core_hwid = CPU_CORE(pnv_core)->core_id;
905 
906         for (j = 0; j < CPU_CORE(pnv_core)->nr_threads; j++) {
907             uint32_t pir = pcc->core_pir(chip, core_hwid) + j;
908             PnvICPState *icp = PNV_ICP(xics_icp_get(xi, pir));
909 
910             memory_region_add_subregion(&chip8->icp_mmio, pir << 12,
911                                         &icp->mmio);
912         }
913     }
914 }
915 
916 static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
917 {
918     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(dev);
919     PnvChip *chip = PNV_CHIP(dev);
920     Pnv8Chip *chip8 = PNV8_CHIP(dev);
921     Pnv8Psi *psi8 = &chip8->psi;
922     Error *local_err = NULL;
923 
924     /* XSCOM bridge is first */
925     pnv_xscom_realize(chip, PNV_XSCOM_SIZE, &local_err);
926     if (local_err) {
927         error_propagate(errp, local_err);
928         return;
929     }
930     sysbus_mmio_map(SYS_BUS_DEVICE(chip), 0, PNV_XSCOM_BASE(chip));
931 
932     pcc->parent_realize(dev, &local_err);
933     if (local_err) {
934         error_propagate(errp, local_err);
935         return;
936     }
937 
938     /* Processor Service Interface (PSI) Host Bridge */
939     object_property_set_int(OBJECT(&chip8->psi), PNV_PSIHB_BASE(chip),
940                             "bar", &error_fatal);
941     object_property_set_bool(OBJECT(&chip8->psi), true, "realized", &local_err);
942     if (local_err) {
943         error_propagate(errp, local_err);
944         return;
945     }
946     pnv_xscom_add_subregion(chip, PNV_XSCOM_PSIHB_BASE,
947                             &PNV_PSI(psi8)->xscom_regs);
948 
949     /* Create LPC controller */
950     object_property_set_bool(OBJECT(&chip8->lpc), true, "realized",
951                              &error_fatal);
952     pnv_xscom_add_subregion(chip, PNV_XSCOM_LPC_BASE, &chip8->lpc.xscom_regs);
953 
954     chip->dt_isa_nodename = g_strdup_printf("/xscom@%" PRIx64 "/isa@%x",
955                                             (uint64_t) PNV_XSCOM_BASE(chip),
956                                             PNV_XSCOM_LPC_BASE);
957 
958     /*
959      * Interrupt Management Area. This is the memory region holding
960      * all the Interrupt Control Presenter (ICP) registers
961      */
962     pnv_chip_icp_realize(chip8, &local_err);
963     if (local_err) {
964         error_propagate(errp, local_err);
965         return;
966     }
967 
968     /* Create the simplified OCC model */
969     object_property_set_bool(OBJECT(&chip8->occ), true, "realized", &local_err);
970     if (local_err) {
971         error_propagate(errp, local_err);
972         return;
973     }
974     pnv_xscom_add_subregion(chip, PNV_XSCOM_OCC_BASE, &chip8->occ.xscom_regs);
975 
976     /* OCC SRAM model */
977     memory_region_add_subregion(get_system_memory(), PNV_OCC_COMMON_AREA(chip),
978                                 &chip8->occ.sram_regs);
979 
980     /* HOMER */
981     object_property_set_bool(OBJECT(&chip8->homer), true, "realized",
982                              &local_err);
983     if (local_err) {
984         error_propagate(errp, local_err);
985         return;
986     }
987     memory_region_add_subregion(get_system_memory(), PNV_HOMER_BASE(chip),
988                                 &chip8->homer.regs);
989 }
990 
991 static void pnv_chip_power8e_class_init(ObjectClass *klass, void *data)
992 {
993     DeviceClass *dc = DEVICE_CLASS(klass);
994     PnvChipClass *k = PNV_CHIP_CLASS(klass);
995 
996     k->chip_type = PNV_CHIP_POWER8E;
997     k->chip_cfam_id = 0x221ef04980000000ull;  /* P8 Murano DD2.1 */
998     k->cores_mask = POWER8E_CORE_MASK;
999     k->core_pir = pnv_chip_core_pir_p8;
1000     k->intc_create = pnv_chip_power8_intc_create;
1001     k->intc_reset = pnv_chip_power8_intc_reset;
1002     k->isa_create = pnv_chip_power8_isa_create;
1003     k->dt_populate = pnv_chip_power8_dt_populate;
1004     k->pic_print_info = pnv_chip_power8_pic_print_info;
1005     dc->desc = "PowerNV Chip POWER8E";
1006 
1007     device_class_set_parent_realize(dc, pnv_chip_power8_realize,
1008                                     &k->parent_realize);
1009 }
1010 
1011 static void pnv_chip_power8_class_init(ObjectClass *klass, void *data)
1012 {
1013     DeviceClass *dc = DEVICE_CLASS(klass);
1014     PnvChipClass *k = PNV_CHIP_CLASS(klass);
1015 
1016     k->chip_type = PNV_CHIP_POWER8;
1017     k->chip_cfam_id = 0x220ea04980000000ull; /* P8 Venice DD2.0 */
1018     k->cores_mask = POWER8_CORE_MASK;
1019     k->core_pir = pnv_chip_core_pir_p8;
1020     k->intc_create = pnv_chip_power8_intc_create;
1021     k->intc_reset = pnv_chip_power8_intc_reset;
1022     k->isa_create = pnv_chip_power8_isa_create;
1023     k->dt_populate = pnv_chip_power8_dt_populate;
1024     k->pic_print_info = pnv_chip_power8_pic_print_info;
1025     dc->desc = "PowerNV Chip POWER8";
1026 
1027     device_class_set_parent_realize(dc, pnv_chip_power8_realize,
1028                                     &k->parent_realize);
1029 }
1030 
1031 static void pnv_chip_power8nvl_class_init(ObjectClass *klass, void *data)
1032 {
1033     DeviceClass *dc = DEVICE_CLASS(klass);
1034     PnvChipClass *k = PNV_CHIP_CLASS(klass);
1035 
1036     k->chip_type = PNV_CHIP_POWER8NVL;
1037     k->chip_cfam_id = 0x120d304980000000ull;  /* P8 Naples DD1.0 */
1038     k->cores_mask = POWER8_CORE_MASK;
1039     k->core_pir = pnv_chip_core_pir_p8;
1040     k->intc_create = pnv_chip_power8_intc_create;
1041     k->intc_reset = pnv_chip_power8_intc_reset;
1042     k->isa_create = pnv_chip_power8nvl_isa_create;
1043     k->dt_populate = pnv_chip_power8_dt_populate;
1044     k->pic_print_info = pnv_chip_power8_pic_print_info;
1045     dc->desc = "PowerNV Chip POWER8NVL";
1046 
1047     device_class_set_parent_realize(dc, pnv_chip_power8_realize,
1048                                     &k->parent_realize);
1049 }
1050 
1051 static void pnv_chip_power9_instance_init(Object *obj)
1052 {
1053     Pnv9Chip *chip9 = PNV9_CHIP(obj);
1054 
1055     object_initialize_child(obj, "xive", &chip9->xive, sizeof(chip9->xive),
1056                             TYPE_PNV_XIVE, &error_abort, NULL);
1057     object_property_add_const_link(OBJECT(&chip9->xive), "chip", obj,
1058                                    &error_abort);
1059 
1060     object_initialize_child(obj, "psi",  &chip9->psi, sizeof(chip9->psi),
1061                             TYPE_PNV9_PSI, &error_abort, NULL);
1062     object_property_add_const_link(OBJECT(&chip9->psi), "chip", obj,
1063                                    &error_abort);
1064 
1065     object_initialize_child(obj, "lpc",  &chip9->lpc, sizeof(chip9->lpc),
1066                             TYPE_PNV9_LPC, &error_abort, NULL);
1067     object_property_add_const_link(OBJECT(&chip9->lpc), "psi",
1068                                    OBJECT(&chip9->psi), &error_abort);
1069 
1070     object_initialize_child(obj, "occ",  &chip9->occ, sizeof(chip9->occ),
1071                             TYPE_PNV9_OCC, &error_abort, NULL);
1072     object_property_add_const_link(OBJECT(&chip9->occ), "psi",
1073                                    OBJECT(&chip9->psi), &error_abort);
1074 
1075     object_initialize_child(obj, "homer",  &chip9->homer, sizeof(chip9->homer),
1076                             TYPE_PNV9_HOMER, &error_abort, NULL);
1077     object_property_add_const_link(OBJECT(&chip9->homer), "chip", obj,
1078                                    &error_abort);
1079 }
1080 
1081 static void pnv_chip_quad_realize(Pnv9Chip *chip9, Error **errp)
1082 {
1083     PnvChip *chip = PNV_CHIP(chip9);
1084     const char *typename = pnv_chip_core_typename(chip);
1085     size_t typesize = object_type_get_instance_size(typename);
1086     int i;
1087 
1088     chip9->nr_quads = DIV_ROUND_UP(chip->nr_cores, 4);
1089     chip9->quads = g_new0(PnvQuad, chip9->nr_quads);
1090 
1091     for (i = 0; i < chip9->nr_quads; i++) {
1092         char eq_name[32];
1093         PnvQuad *eq = &chip9->quads[i];
1094         PnvCore *pnv_core = PNV_CORE(chip->cores + (i * 4) * typesize);
1095         int core_id = CPU_CORE(pnv_core)->core_id;
1096 
1097         snprintf(eq_name, sizeof(eq_name), "eq[%d]", core_id);
1098         object_initialize_child(OBJECT(chip), eq_name, eq, sizeof(*eq),
1099                                 TYPE_PNV_QUAD, &error_fatal, NULL);
1100 
1101         object_property_set_int(OBJECT(eq), core_id, "id", &error_fatal);
1102         object_property_set_bool(OBJECT(eq), true, "realized", &error_fatal);
1103 
1104         pnv_xscom_add_subregion(chip, PNV9_XSCOM_EQ_BASE(eq->id),
1105                                 &eq->xscom_regs);
1106     }
1107 }
1108 
1109 static void pnv_chip_power9_realize(DeviceState *dev, Error **errp)
1110 {
1111     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(dev);
1112     Pnv9Chip *chip9 = PNV9_CHIP(dev);
1113     PnvChip *chip = PNV_CHIP(dev);
1114     Pnv9Psi *psi9 = &chip9->psi;
1115     Error *local_err = NULL;
1116 
1117     /* XSCOM bridge is first */
1118     pnv_xscom_realize(chip, PNV9_XSCOM_SIZE, &local_err);
1119     if (local_err) {
1120         error_propagate(errp, local_err);
1121         return;
1122     }
1123     sysbus_mmio_map(SYS_BUS_DEVICE(chip), 0, PNV9_XSCOM_BASE(chip));
1124 
1125     pcc->parent_realize(dev, &local_err);
1126     if (local_err) {
1127         error_propagate(errp, local_err);
1128         return;
1129     }
1130 
1131     pnv_chip_quad_realize(chip9, &local_err);
1132     if (local_err) {
1133         error_propagate(errp, local_err);
1134         return;
1135     }
1136 
1137     /* XIVE interrupt controller (POWER9) */
1138     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_IC_BASE(chip),
1139                             "ic-bar", &error_fatal);
1140     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_VC_BASE(chip),
1141                             "vc-bar", &error_fatal);
1142     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_PC_BASE(chip),
1143                             "pc-bar", &error_fatal);
1144     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_TM_BASE(chip),
1145                             "tm-bar", &error_fatal);
1146     object_property_set_bool(OBJECT(&chip9->xive), true, "realized",
1147                              &local_err);
1148     if (local_err) {
1149         error_propagate(errp, local_err);
1150         return;
1151     }
1152     pnv_xscom_add_subregion(chip, PNV9_XSCOM_XIVE_BASE,
1153                             &chip9->xive.xscom_regs);
1154 
1155     /* Processor Service Interface (PSI) Host Bridge */
1156     object_property_set_int(OBJECT(&chip9->psi), PNV9_PSIHB_BASE(chip),
1157                             "bar", &error_fatal);
1158     object_property_set_bool(OBJECT(&chip9->psi), true, "realized", &local_err);
1159     if (local_err) {
1160         error_propagate(errp, local_err);
1161         return;
1162     }
1163     pnv_xscom_add_subregion(chip, PNV9_XSCOM_PSIHB_BASE,
1164                             &PNV_PSI(psi9)->xscom_regs);
1165 
1166     /* LPC */
1167     object_property_set_bool(OBJECT(&chip9->lpc), true, "realized", &local_err);
1168     if (local_err) {
1169         error_propagate(errp, local_err);
1170         return;
1171     }
1172     memory_region_add_subregion(get_system_memory(), PNV9_LPCM_BASE(chip),
1173                                 &chip9->lpc.xscom_regs);
1174 
1175     chip->dt_isa_nodename = g_strdup_printf("/lpcm-opb@%" PRIx64 "/lpc@0",
1176                                             (uint64_t) PNV9_LPCM_BASE(chip));
1177 
1178     /* Create the simplified OCC model */
1179     object_property_set_bool(OBJECT(&chip9->occ), true, "realized", &local_err);
1180     if (local_err) {
1181         error_propagate(errp, local_err);
1182         return;
1183     }
1184     pnv_xscom_add_subregion(chip, PNV9_XSCOM_OCC_BASE, &chip9->occ.xscom_regs);
1185 
1186     /* OCC SRAM model */
1187     memory_region_add_subregion(get_system_memory(), PNV9_OCC_COMMON_AREA(chip),
1188                                 &chip9->occ.sram_regs);
1189 
1190     /* HOMER */
1191     object_property_set_bool(OBJECT(&chip9->homer), true, "realized",
1192                              &local_err);
1193     if (local_err) {
1194         error_propagate(errp, local_err);
1195         return;
1196     }
1197     memory_region_add_subregion(get_system_memory(), PNV9_HOMER_BASE(chip),
1198                                 &chip9->homer.regs);
1199 }
1200 
1201 static void pnv_chip_power9_class_init(ObjectClass *klass, void *data)
1202 {
1203     DeviceClass *dc = DEVICE_CLASS(klass);
1204     PnvChipClass *k = PNV_CHIP_CLASS(klass);
1205 
1206     k->chip_type = PNV_CHIP_POWER9;
1207     k->chip_cfam_id = 0x220d104900008000ull; /* P9 Nimbus DD2.0 */
1208     k->cores_mask = POWER9_CORE_MASK;
1209     k->core_pir = pnv_chip_core_pir_p9;
1210     k->intc_create = pnv_chip_power9_intc_create;
1211     k->intc_reset = pnv_chip_power9_intc_reset;
1212     k->isa_create = pnv_chip_power9_isa_create;
1213     k->dt_populate = pnv_chip_power9_dt_populate;
1214     k->pic_print_info = pnv_chip_power9_pic_print_info;
1215     dc->desc = "PowerNV Chip POWER9";
1216 
1217     device_class_set_parent_realize(dc, pnv_chip_power9_realize,
1218                                     &k->parent_realize);
1219 }
1220 
1221 static void pnv_chip_core_sanitize(PnvChip *chip, Error **errp)
1222 {
1223     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
1224     int cores_max;
1225 
1226     /*
1227      * No custom mask for this chip, let's use the default one from *
1228      * the chip class
1229      */
1230     if (!chip->cores_mask) {
1231         chip->cores_mask = pcc->cores_mask;
1232     }
1233 
1234     /* filter alien core ids ! some are reserved */
1235     if ((chip->cores_mask & pcc->cores_mask) != chip->cores_mask) {
1236         error_setg(errp, "warning: invalid core mask for chip Ox%"PRIx64" !",
1237                    chip->cores_mask);
1238         return;
1239     }
1240     chip->cores_mask &= pcc->cores_mask;
1241 
1242     /* now that we have a sane layout, let check the number of cores */
1243     cores_max = ctpop64(chip->cores_mask);
1244     if (chip->nr_cores > cores_max) {
1245         error_setg(errp, "warning: too many cores for chip ! Limit is %d",
1246                    cores_max);
1247         return;
1248     }
1249 }
1250 
1251 static void pnv_chip_core_realize(PnvChip *chip, Error **errp)
1252 {
1253     MachineState *ms = MACHINE(qdev_get_machine());
1254     Error *error = NULL;
1255     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
1256     const char *typename = pnv_chip_core_typename(chip);
1257     size_t typesize = object_type_get_instance_size(typename);
1258     int i, core_hwid;
1259 
1260     if (!object_class_by_name(typename)) {
1261         error_setg(errp, "Unable to find PowerNV CPU Core '%s'", typename);
1262         return;
1263     }
1264 
1265     /* Cores */
1266     pnv_chip_core_sanitize(chip, &error);
1267     if (error) {
1268         error_propagate(errp, error);
1269         return;
1270     }
1271 
1272     chip->cores = g_malloc0(typesize * chip->nr_cores);
1273 
1274     for (i = 0, core_hwid = 0; (core_hwid < sizeof(chip->cores_mask) * 8)
1275              && (i < chip->nr_cores); core_hwid++) {
1276         char core_name[32];
1277         void *pnv_core = chip->cores + i * typesize;
1278         uint64_t xscom_core_base;
1279 
1280         if (!(chip->cores_mask & (1ull << core_hwid))) {
1281             continue;
1282         }
1283 
1284         snprintf(core_name, sizeof(core_name), "core[%d]", core_hwid);
1285         object_initialize_child(OBJECT(chip), core_name, pnv_core, typesize,
1286                                 typename, &error_fatal, NULL);
1287         object_property_set_int(OBJECT(pnv_core), ms->smp.threads, "nr-threads",
1288                                 &error_fatal);
1289         object_property_set_int(OBJECT(pnv_core), core_hwid,
1290                                 CPU_CORE_PROP_CORE_ID, &error_fatal);
1291         object_property_set_int(OBJECT(pnv_core),
1292                                 pcc->core_pir(chip, core_hwid),
1293                                 "pir", &error_fatal);
1294         object_property_add_const_link(OBJECT(pnv_core), "chip",
1295                                        OBJECT(chip), &error_fatal);
1296         object_property_set_bool(OBJECT(pnv_core), true, "realized",
1297                                  &error_fatal);
1298 
1299         /* Each core has an XSCOM MMIO region */
1300         if (!pnv_chip_is_power9(chip)) {
1301             xscom_core_base = PNV_XSCOM_EX_BASE(core_hwid);
1302         } else {
1303             xscom_core_base = PNV9_XSCOM_EC_BASE(core_hwid);
1304         }
1305 
1306         pnv_xscom_add_subregion(chip, xscom_core_base,
1307                                 &PNV_CORE(pnv_core)->xscom_regs);
1308         i++;
1309     }
1310 }
1311 
1312 static void pnv_chip_realize(DeviceState *dev, Error **errp)
1313 {
1314     PnvChip *chip = PNV_CHIP(dev);
1315     Error *error = NULL;
1316 
1317     /* Cores */
1318     pnv_chip_core_realize(chip, &error);
1319     if (error) {
1320         error_propagate(errp, error);
1321         return;
1322     }
1323 }
1324 
1325 static Property pnv_chip_properties[] = {
1326     DEFINE_PROP_UINT32("chip-id", PnvChip, chip_id, 0),
1327     DEFINE_PROP_UINT64("ram-start", PnvChip, ram_start, 0),
1328     DEFINE_PROP_UINT64("ram-size", PnvChip, ram_size, 0),
1329     DEFINE_PROP_UINT32("nr-cores", PnvChip, nr_cores, 1),
1330     DEFINE_PROP_UINT64("cores-mask", PnvChip, cores_mask, 0x0),
1331     DEFINE_PROP_END_OF_LIST(),
1332 };
1333 
1334 static void pnv_chip_class_init(ObjectClass *klass, void *data)
1335 {
1336     DeviceClass *dc = DEVICE_CLASS(klass);
1337 
1338     set_bit(DEVICE_CATEGORY_CPU, dc->categories);
1339     dc->realize = pnv_chip_realize;
1340     dc->props = pnv_chip_properties;
1341     dc->desc = "PowerNV Chip";
1342 }
1343 
1344 static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
1345 {
1346     PnvMachineState *pnv = PNV_MACHINE(xi);
1347     int i;
1348 
1349     for (i = 0; i < pnv->num_chips; i++) {
1350         Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
1351 
1352         if (ics_valid_irq(&chip8->psi.ics, irq)) {
1353             return &chip8->psi.ics;
1354         }
1355     }
1356     return NULL;
1357 }
1358 
1359 static void pnv_ics_resend(XICSFabric *xi)
1360 {
1361     PnvMachineState *pnv = PNV_MACHINE(xi);
1362     int i;
1363 
1364     for (i = 0; i < pnv->num_chips; i++) {
1365         Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
1366         ics_resend(&chip8->psi.ics);
1367     }
1368 }
1369 
1370 static ICPState *pnv_icp_get(XICSFabric *xi, int pir)
1371 {
1372     PowerPCCPU *cpu = ppc_get_vcpu_by_pir(pir);
1373 
1374     return cpu ? ICP(pnv_cpu_state(cpu)->intc) : NULL;
1375 }
1376 
1377 static void pnv_pic_print_info(InterruptStatsProvider *obj,
1378                                Monitor *mon)
1379 {
1380     PnvMachineState *pnv = PNV_MACHINE(obj);
1381     int i;
1382     CPUState *cs;
1383 
1384     CPU_FOREACH(cs) {
1385         PowerPCCPU *cpu = POWERPC_CPU(cs);
1386 
1387         if (pnv_chip_is_power9(pnv->chips[0])) {
1388             xive_tctx_pic_print_info(XIVE_TCTX(pnv_cpu_state(cpu)->intc), mon);
1389         } else {
1390             icp_pic_print_info(ICP(pnv_cpu_state(cpu)->intc), mon);
1391         }
1392     }
1393 
1394     for (i = 0; i < pnv->num_chips; i++) {
1395         PNV_CHIP_GET_CLASS(pnv->chips[i])->pic_print_info(pnv->chips[i], mon);
1396     }
1397 }
1398 
1399 static void pnv_get_num_chips(Object *obj, Visitor *v, const char *name,
1400                               void *opaque, Error **errp)
1401 {
1402     visit_type_uint32(v, name, &PNV_MACHINE(obj)->num_chips, errp);
1403 }
1404 
1405 static void pnv_set_num_chips(Object *obj, Visitor *v, const char *name,
1406                               void *opaque, Error **errp)
1407 {
1408     PnvMachineState *pnv = PNV_MACHINE(obj);
1409     uint32_t num_chips;
1410     Error *local_err = NULL;
1411 
1412     visit_type_uint32(v, name, &num_chips, &local_err);
1413     if (local_err) {
1414         error_propagate(errp, local_err);
1415         return;
1416     }
1417 
1418     /*
1419      * TODO: should we decide on how many chips we can create based
1420      * on #cores and Venice vs. Murano vs. Naples chip type etc...,
1421      */
1422     if (!is_power_of_2(num_chips) || num_chips > 4) {
1423         error_setg(errp, "invalid number of chips: '%d'", num_chips);
1424         return;
1425     }
1426 
1427     pnv->num_chips = num_chips;
1428 }
1429 
1430 static void pnv_machine_instance_init(Object *obj)
1431 {
1432     PnvMachineState *pnv = PNV_MACHINE(obj);
1433     pnv->num_chips = 1;
1434 }
1435 
1436 static void pnv_machine_class_props_init(ObjectClass *oc)
1437 {
1438     object_class_property_add(oc, "num-chips", "uint32",
1439                               pnv_get_num_chips, pnv_set_num_chips,
1440                               NULL, NULL, NULL);
1441     object_class_property_set_description(oc, "num-chips",
1442                               "Specifies the number of processor chips",
1443                               NULL);
1444 }
1445 
1446 static void pnv_machine_power8_class_init(ObjectClass *oc, void *data)
1447 {
1448     MachineClass *mc = MACHINE_CLASS(oc);
1449     XICSFabricClass *xic = XICS_FABRIC_CLASS(oc);
1450 
1451     mc->desc = "IBM PowerNV (Non-Virtualized) POWER8";
1452     mc->default_cpu_type = POWERPC_CPU_TYPE_NAME("power8_v2.0");
1453 
1454     xic->icp_get = pnv_icp_get;
1455     xic->ics_get = pnv_ics_get;
1456     xic->ics_resend = pnv_ics_resend;
1457 }
1458 
1459 static void pnv_machine_power9_class_init(ObjectClass *oc, void *data)
1460 {
1461     MachineClass *mc = MACHINE_CLASS(oc);
1462 
1463     mc->desc = "IBM PowerNV (Non-Virtualized) POWER9";
1464     mc->default_cpu_type = POWERPC_CPU_TYPE_NAME("power9_v2.0");
1465 
1466     mc->alias = "powernv";
1467 }
1468 
1469 static void pnv_machine_class_init(ObjectClass *oc, void *data)
1470 {
1471     MachineClass *mc = MACHINE_CLASS(oc);
1472     InterruptStatsProviderClass *ispc = INTERRUPT_STATS_PROVIDER_CLASS(oc);
1473 
1474     mc->desc = "IBM PowerNV (Non-Virtualized)";
1475     mc->init = pnv_init;
1476     mc->reset = pnv_reset;
1477     mc->max_cpus = MAX_CPUS;
1478     /* Pnv provides a AHCI device for storage */
1479     mc->block_default_type = IF_IDE;
1480     mc->no_parallel = 1;
1481     mc->default_boot_order = NULL;
1482     /*
1483      * RAM defaults to less than 2048 for 32-bit hosts, and large
1484      * enough to fit the maximum initrd size at it's load address
1485      */
1486     mc->default_ram_size = INITRD_LOAD_ADDR + INITRD_MAX_SIZE;
1487     ispc->print_info = pnv_pic_print_info;
1488 
1489     pnv_machine_class_props_init(oc);
1490 }
1491 
1492 #define DEFINE_PNV8_CHIP_TYPE(type, class_initfn) \
1493     {                                             \
1494         .name          = type,                    \
1495         .class_init    = class_initfn,            \
1496         .parent        = TYPE_PNV8_CHIP,          \
1497     }
1498 
1499 #define DEFINE_PNV9_CHIP_TYPE(type, class_initfn) \
1500     {                                             \
1501         .name          = type,                    \
1502         .class_init    = class_initfn,            \
1503         .parent        = TYPE_PNV9_CHIP,          \
1504     }
1505 
1506 static const TypeInfo types[] = {
1507     {
1508         .name          = MACHINE_TYPE_NAME("powernv9"),
1509         .parent        = TYPE_PNV_MACHINE,
1510         .class_init    = pnv_machine_power9_class_init,
1511     },
1512     {
1513         .name          = MACHINE_TYPE_NAME("powernv8"),
1514         .parent        = TYPE_PNV_MACHINE,
1515         .class_init    = pnv_machine_power8_class_init,
1516         .interfaces = (InterfaceInfo[]) {
1517             { TYPE_XICS_FABRIC },
1518             { },
1519         },
1520     },
1521     {
1522         .name          = TYPE_PNV_MACHINE,
1523         .parent        = TYPE_MACHINE,
1524         .abstract       = true,
1525         .instance_size = sizeof(PnvMachineState),
1526         .instance_init = pnv_machine_instance_init,
1527         .class_init    = pnv_machine_class_init,
1528         .interfaces = (InterfaceInfo[]) {
1529             { TYPE_INTERRUPT_STATS_PROVIDER },
1530             { },
1531         },
1532     },
1533     {
1534         .name          = TYPE_PNV_CHIP,
1535         .parent        = TYPE_SYS_BUS_DEVICE,
1536         .class_init    = pnv_chip_class_init,
1537         .instance_size = sizeof(PnvChip),
1538         .class_size    = sizeof(PnvChipClass),
1539         .abstract      = true,
1540     },
1541 
1542     /*
1543      * P9 chip and variants
1544      */
1545     {
1546         .name          = TYPE_PNV9_CHIP,
1547         .parent        = TYPE_PNV_CHIP,
1548         .instance_init = pnv_chip_power9_instance_init,
1549         .instance_size = sizeof(Pnv9Chip),
1550     },
1551     DEFINE_PNV9_CHIP_TYPE(TYPE_PNV_CHIP_POWER9, pnv_chip_power9_class_init),
1552 
1553     /*
1554      * P8 chip and variants
1555      */
1556     {
1557         .name          = TYPE_PNV8_CHIP,
1558         .parent        = TYPE_PNV_CHIP,
1559         .instance_init = pnv_chip_power8_instance_init,
1560         .instance_size = sizeof(Pnv8Chip),
1561     },
1562     DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8, pnv_chip_power8_class_init),
1563     DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8E, pnv_chip_power8e_class_init),
1564     DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8NVL,
1565                           pnv_chip_power8nvl_class_init),
1566 };
1567 
1568 DEFINE_TYPES(types)
1569