xref: /openbmc/qemu/hw/ppc/pnv.c (revision e4ec5ad4)
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/timer/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 /*
782  *    0:48  Reserved - Read as zeroes
783  *   49:52  Node ID
784  *   53:55  Chip ID
785  *   56     Reserved - Read as zero
786  *   57:61  Core number
787  *   62:63  Thread ID
788  *
789  * We only care about the lower bits. uint32_t is fine for the moment.
790  */
791 static uint32_t pnv_chip_core_pir_p9(PnvChip *chip, uint32_t core_id)
792 {
793     return (chip->chip_id << 8) | (core_id << 2);
794 }
795 
796 static void pnv_chip_power9_intc_create(PnvChip *chip, PowerPCCPU *cpu,
797                                         Error **errp)
798 {
799     Pnv9Chip *chip9 = PNV9_CHIP(chip);
800     Error *local_err = NULL;
801     Object *obj;
802     PnvCPUState *pnv_cpu = pnv_cpu_state(cpu);
803 
804     /*
805      * The core creates its interrupt presenter but the XIVE interrupt
806      * controller object is initialized afterwards. Hopefully, it's
807      * only used at runtime.
808      */
809     obj = xive_tctx_create(OBJECT(cpu), XIVE_ROUTER(&chip9->xive), &local_err);
810     if (local_err) {
811         error_propagate(errp, local_err);
812         return;
813     }
814 
815     pnv_cpu->intc = obj;
816 }
817 
818 /*
819  * Allowed core identifiers on a POWER8 Processor Chip :
820  *
821  * <EX0 reserved>
822  *  EX1  - Venice only
823  *  EX2  - Venice only
824  *  EX3  - Venice only
825  *  EX4
826  *  EX5
827  *  EX6
828  * <EX7,8 reserved> <reserved>
829  *  EX9  - Venice only
830  *  EX10 - Venice only
831  *  EX11 - Venice only
832  *  EX12
833  *  EX13
834  *  EX14
835  * <EX15 reserved>
836  */
837 #define POWER8E_CORE_MASK  (0x7070ull)
838 #define POWER8_CORE_MASK   (0x7e7eull)
839 
840 /*
841  * POWER9 has 24 cores, ids starting at 0x0
842  */
843 #define POWER9_CORE_MASK   (0xffffffffffffffull)
844 
845 static void pnv_chip_power8_instance_init(Object *obj)
846 {
847     Pnv8Chip *chip8 = PNV8_CHIP(obj);
848 
849     object_initialize_child(obj, "psi",  &chip8->psi, sizeof(chip8->psi),
850                             TYPE_PNV8_PSI, &error_abort, NULL);
851     object_property_add_const_link(OBJECT(&chip8->psi), "xics",
852                                    OBJECT(qdev_get_machine()), &error_abort);
853 
854     object_initialize_child(obj, "lpc",  &chip8->lpc, sizeof(chip8->lpc),
855                             TYPE_PNV8_LPC, &error_abort, NULL);
856     object_property_add_const_link(OBJECT(&chip8->lpc), "psi",
857                                    OBJECT(&chip8->psi), &error_abort);
858 
859     object_initialize_child(obj, "occ",  &chip8->occ, sizeof(chip8->occ),
860                             TYPE_PNV8_OCC, &error_abort, NULL);
861     object_property_add_const_link(OBJECT(&chip8->occ), "psi",
862                                    OBJECT(&chip8->psi), &error_abort);
863 
864     object_initialize_child(obj, "homer",  &chip8->homer, sizeof(chip8->homer),
865                             TYPE_PNV8_HOMER, &error_abort, NULL);
866     object_property_add_const_link(OBJECT(&chip8->homer), "chip", obj,
867                                    &error_abort);
868 }
869 
870 static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
871  {
872     PnvChip *chip = PNV_CHIP(chip8);
873     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
874     const char *typename = pnv_chip_core_typename(chip);
875     size_t typesize = object_type_get_instance_size(typename);
876     int i, j;
877     char *name;
878     XICSFabric *xi = XICS_FABRIC(qdev_get_machine());
879 
880     name = g_strdup_printf("icp-%x", chip->chip_id);
881     memory_region_init(&chip8->icp_mmio, OBJECT(chip), name, PNV_ICP_SIZE);
882     sysbus_init_mmio(SYS_BUS_DEVICE(chip), &chip8->icp_mmio);
883     g_free(name);
884 
885     sysbus_mmio_map(SYS_BUS_DEVICE(chip), 1, PNV_ICP_BASE(chip));
886 
887     /* Map the ICP registers for each thread */
888     for (i = 0; i < chip->nr_cores; i++) {
889         PnvCore *pnv_core = PNV_CORE(chip->cores + i * typesize);
890         int core_hwid = CPU_CORE(pnv_core)->core_id;
891 
892         for (j = 0; j < CPU_CORE(pnv_core)->nr_threads; j++) {
893             uint32_t pir = pcc->core_pir(chip, core_hwid) + j;
894             PnvICPState *icp = PNV_ICP(xics_icp_get(xi, pir));
895 
896             memory_region_add_subregion(&chip8->icp_mmio, pir << 12,
897                                         &icp->mmio);
898         }
899     }
900 }
901 
902 static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
903 {
904     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(dev);
905     PnvChip *chip = PNV_CHIP(dev);
906     Pnv8Chip *chip8 = PNV8_CHIP(dev);
907     Pnv8Psi *psi8 = &chip8->psi;
908     Error *local_err = NULL;
909 
910     /* XSCOM bridge is first */
911     pnv_xscom_realize(chip, PNV_XSCOM_SIZE, &local_err);
912     if (local_err) {
913         error_propagate(errp, local_err);
914         return;
915     }
916     sysbus_mmio_map(SYS_BUS_DEVICE(chip), 0, PNV_XSCOM_BASE(chip));
917 
918     pcc->parent_realize(dev, &local_err);
919     if (local_err) {
920         error_propagate(errp, local_err);
921         return;
922     }
923 
924     /* Processor Service Interface (PSI) Host Bridge */
925     object_property_set_int(OBJECT(&chip8->psi), PNV_PSIHB_BASE(chip),
926                             "bar", &error_fatal);
927     object_property_set_bool(OBJECT(&chip8->psi), true, "realized", &local_err);
928     if (local_err) {
929         error_propagate(errp, local_err);
930         return;
931     }
932     pnv_xscom_add_subregion(chip, PNV_XSCOM_PSIHB_BASE,
933                             &PNV_PSI(psi8)->xscom_regs);
934 
935     /* Create LPC controller */
936     object_property_set_bool(OBJECT(&chip8->lpc), true, "realized",
937                              &error_fatal);
938     pnv_xscom_add_subregion(chip, PNV_XSCOM_LPC_BASE, &chip8->lpc.xscom_regs);
939 
940     chip->dt_isa_nodename = g_strdup_printf("/xscom@%" PRIx64 "/isa@%x",
941                                             (uint64_t) PNV_XSCOM_BASE(chip),
942                                             PNV_XSCOM_LPC_BASE);
943 
944     /*
945      * Interrupt Management Area. This is the memory region holding
946      * all the Interrupt Control Presenter (ICP) registers
947      */
948     pnv_chip_icp_realize(chip8, &local_err);
949     if (local_err) {
950         error_propagate(errp, local_err);
951         return;
952     }
953 
954     /* Create the simplified OCC model */
955     object_property_set_bool(OBJECT(&chip8->occ), true, "realized", &local_err);
956     if (local_err) {
957         error_propagate(errp, local_err);
958         return;
959     }
960     pnv_xscom_add_subregion(chip, PNV_XSCOM_OCC_BASE, &chip8->occ.xscom_regs);
961 
962     /* OCC SRAM model */
963     memory_region_add_subregion(get_system_memory(), PNV_OCC_COMMON_AREA(chip),
964                                 &chip8->occ.sram_regs);
965 
966     /* HOMER */
967     object_property_set_bool(OBJECT(&chip8->homer), true, "realized",
968                              &local_err);
969     if (local_err) {
970         error_propagate(errp, local_err);
971         return;
972     }
973     memory_region_add_subregion(get_system_memory(), PNV_HOMER_BASE(chip),
974                                 &chip8->homer.regs);
975 }
976 
977 static void pnv_chip_power8e_class_init(ObjectClass *klass, void *data)
978 {
979     DeviceClass *dc = DEVICE_CLASS(klass);
980     PnvChipClass *k = PNV_CHIP_CLASS(klass);
981 
982     k->chip_type = PNV_CHIP_POWER8E;
983     k->chip_cfam_id = 0x221ef04980000000ull;  /* P8 Murano DD2.1 */
984     k->cores_mask = POWER8E_CORE_MASK;
985     k->core_pir = pnv_chip_core_pir_p8;
986     k->intc_create = pnv_chip_power8_intc_create;
987     k->isa_create = pnv_chip_power8_isa_create;
988     k->dt_populate = pnv_chip_power8_dt_populate;
989     k->pic_print_info = pnv_chip_power8_pic_print_info;
990     dc->desc = "PowerNV Chip POWER8E";
991 
992     device_class_set_parent_realize(dc, pnv_chip_power8_realize,
993                                     &k->parent_realize);
994 }
995 
996 static void pnv_chip_power8_class_init(ObjectClass *klass, void *data)
997 {
998     DeviceClass *dc = DEVICE_CLASS(klass);
999     PnvChipClass *k = PNV_CHIP_CLASS(klass);
1000 
1001     k->chip_type = PNV_CHIP_POWER8;
1002     k->chip_cfam_id = 0x220ea04980000000ull; /* P8 Venice DD2.0 */
1003     k->cores_mask = POWER8_CORE_MASK;
1004     k->core_pir = pnv_chip_core_pir_p8;
1005     k->intc_create = pnv_chip_power8_intc_create;
1006     k->isa_create = pnv_chip_power8_isa_create;
1007     k->dt_populate = pnv_chip_power8_dt_populate;
1008     k->pic_print_info = pnv_chip_power8_pic_print_info;
1009     dc->desc = "PowerNV Chip POWER8";
1010 
1011     device_class_set_parent_realize(dc, pnv_chip_power8_realize,
1012                                     &k->parent_realize);
1013 }
1014 
1015 static void pnv_chip_power8nvl_class_init(ObjectClass *klass, void *data)
1016 {
1017     DeviceClass *dc = DEVICE_CLASS(klass);
1018     PnvChipClass *k = PNV_CHIP_CLASS(klass);
1019 
1020     k->chip_type = PNV_CHIP_POWER8NVL;
1021     k->chip_cfam_id = 0x120d304980000000ull;  /* P8 Naples DD1.0 */
1022     k->cores_mask = POWER8_CORE_MASK;
1023     k->core_pir = pnv_chip_core_pir_p8;
1024     k->intc_create = pnv_chip_power8_intc_create;
1025     k->isa_create = pnv_chip_power8nvl_isa_create;
1026     k->dt_populate = pnv_chip_power8_dt_populate;
1027     k->pic_print_info = pnv_chip_power8_pic_print_info;
1028     dc->desc = "PowerNV Chip POWER8NVL";
1029 
1030     device_class_set_parent_realize(dc, pnv_chip_power8_realize,
1031                                     &k->parent_realize);
1032 }
1033 
1034 static void pnv_chip_power9_instance_init(Object *obj)
1035 {
1036     Pnv9Chip *chip9 = PNV9_CHIP(obj);
1037 
1038     object_initialize_child(obj, "xive", &chip9->xive, sizeof(chip9->xive),
1039                             TYPE_PNV_XIVE, &error_abort, NULL);
1040     object_property_add_const_link(OBJECT(&chip9->xive), "chip", obj,
1041                                    &error_abort);
1042 
1043     object_initialize_child(obj, "psi",  &chip9->psi, sizeof(chip9->psi),
1044                             TYPE_PNV9_PSI, &error_abort, NULL);
1045     object_property_add_const_link(OBJECT(&chip9->psi), "chip", obj,
1046                                    &error_abort);
1047 
1048     object_initialize_child(obj, "lpc",  &chip9->lpc, sizeof(chip9->lpc),
1049                             TYPE_PNV9_LPC, &error_abort, NULL);
1050     object_property_add_const_link(OBJECT(&chip9->lpc), "psi",
1051                                    OBJECT(&chip9->psi), &error_abort);
1052 
1053     object_initialize_child(obj, "occ",  &chip9->occ, sizeof(chip9->occ),
1054                             TYPE_PNV9_OCC, &error_abort, NULL);
1055     object_property_add_const_link(OBJECT(&chip9->occ), "psi",
1056                                    OBJECT(&chip9->psi), &error_abort);
1057 
1058     object_initialize_child(obj, "homer",  &chip9->homer, sizeof(chip9->homer),
1059                             TYPE_PNV9_HOMER, &error_abort, NULL);
1060     object_property_add_const_link(OBJECT(&chip9->homer), "chip", obj,
1061                                    &error_abort);
1062 }
1063 
1064 static void pnv_chip_quad_realize(Pnv9Chip *chip9, Error **errp)
1065 {
1066     PnvChip *chip = PNV_CHIP(chip9);
1067     const char *typename = pnv_chip_core_typename(chip);
1068     size_t typesize = object_type_get_instance_size(typename);
1069     int i;
1070 
1071     chip9->nr_quads = DIV_ROUND_UP(chip->nr_cores, 4);
1072     chip9->quads = g_new0(PnvQuad, chip9->nr_quads);
1073 
1074     for (i = 0; i < chip9->nr_quads; i++) {
1075         char eq_name[32];
1076         PnvQuad *eq = &chip9->quads[i];
1077         PnvCore *pnv_core = PNV_CORE(chip->cores + (i * 4) * typesize);
1078         int core_id = CPU_CORE(pnv_core)->core_id;
1079 
1080         snprintf(eq_name, sizeof(eq_name), "eq[%d]", core_id);
1081         object_initialize_child(OBJECT(chip), eq_name, eq, sizeof(*eq),
1082                                 TYPE_PNV_QUAD, &error_fatal, NULL);
1083 
1084         object_property_set_int(OBJECT(eq), core_id, "id", &error_fatal);
1085         object_property_set_bool(OBJECT(eq), true, "realized", &error_fatal);
1086 
1087         pnv_xscom_add_subregion(chip, PNV9_XSCOM_EQ_BASE(eq->id),
1088                                 &eq->xscom_regs);
1089     }
1090 }
1091 
1092 static void pnv_chip_power9_realize(DeviceState *dev, Error **errp)
1093 {
1094     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(dev);
1095     Pnv9Chip *chip9 = PNV9_CHIP(dev);
1096     PnvChip *chip = PNV_CHIP(dev);
1097     Pnv9Psi *psi9 = &chip9->psi;
1098     Error *local_err = NULL;
1099 
1100     /* XSCOM bridge is first */
1101     pnv_xscom_realize(chip, PNV9_XSCOM_SIZE, &local_err);
1102     if (local_err) {
1103         error_propagate(errp, local_err);
1104         return;
1105     }
1106     sysbus_mmio_map(SYS_BUS_DEVICE(chip), 0, PNV9_XSCOM_BASE(chip));
1107 
1108     pcc->parent_realize(dev, &local_err);
1109     if (local_err) {
1110         error_propagate(errp, local_err);
1111         return;
1112     }
1113 
1114     pnv_chip_quad_realize(chip9, &local_err);
1115     if (local_err) {
1116         error_propagate(errp, local_err);
1117         return;
1118     }
1119 
1120     /* XIVE interrupt controller (POWER9) */
1121     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_IC_BASE(chip),
1122                             "ic-bar", &error_fatal);
1123     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_VC_BASE(chip),
1124                             "vc-bar", &error_fatal);
1125     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_PC_BASE(chip),
1126                             "pc-bar", &error_fatal);
1127     object_property_set_int(OBJECT(&chip9->xive), PNV9_XIVE_TM_BASE(chip),
1128                             "tm-bar", &error_fatal);
1129     object_property_set_bool(OBJECT(&chip9->xive), true, "realized",
1130                              &local_err);
1131     if (local_err) {
1132         error_propagate(errp, local_err);
1133         return;
1134     }
1135     pnv_xscom_add_subregion(chip, PNV9_XSCOM_XIVE_BASE,
1136                             &chip9->xive.xscom_regs);
1137 
1138     /* Processor Service Interface (PSI) Host Bridge */
1139     object_property_set_int(OBJECT(&chip9->psi), PNV9_PSIHB_BASE(chip),
1140                             "bar", &error_fatal);
1141     object_property_set_bool(OBJECT(&chip9->psi), true, "realized", &local_err);
1142     if (local_err) {
1143         error_propagate(errp, local_err);
1144         return;
1145     }
1146     pnv_xscom_add_subregion(chip, PNV9_XSCOM_PSIHB_BASE,
1147                             &PNV_PSI(psi9)->xscom_regs);
1148 
1149     /* LPC */
1150     object_property_set_bool(OBJECT(&chip9->lpc), true, "realized", &local_err);
1151     if (local_err) {
1152         error_propagate(errp, local_err);
1153         return;
1154     }
1155     memory_region_add_subregion(get_system_memory(), PNV9_LPCM_BASE(chip),
1156                                 &chip9->lpc.xscom_regs);
1157 
1158     chip->dt_isa_nodename = g_strdup_printf("/lpcm-opb@%" PRIx64 "/lpc@0",
1159                                             (uint64_t) PNV9_LPCM_BASE(chip));
1160 
1161     /* Create the simplified OCC model */
1162     object_property_set_bool(OBJECT(&chip9->occ), true, "realized", &local_err);
1163     if (local_err) {
1164         error_propagate(errp, local_err);
1165         return;
1166     }
1167     pnv_xscom_add_subregion(chip, PNV9_XSCOM_OCC_BASE, &chip9->occ.xscom_regs);
1168 
1169     /* OCC SRAM model */
1170     memory_region_add_subregion(get_system_memory(), PNV9_OCC_COMMON_AREA(chip),
1171                                 &chip9->occ.sram_regs);
1172 
1173     /* HOMER */
1174     object_property_set_bool(OBJECT(&chip9->homer), true, "realized",
1175                              &local_err);
1176     if (local_err) {
1177         error_propagate(errp, local_err);
1178         return;
1179     }
1180     memory_region_add_subregion(get_system_memory(), PNV9_HOMER_BASE(chip),
1181                                 &chip9->homer.regs);
1182 }
1183 
1184 static void pnv_chip_power9_class_init(ObjectClass *klass, void *data)
1185 {
1186     DeviceClass *dc = DEVICE_CLASS(klass);
1187     PnvChipClass *k = PNV_CHIP_CLASS(klass);
1188 
1189     k->chip_type = PNV_CHIP_POWER9;
1190     k->chip_cfam_id = 0x220d104900008000ull; /* P9 Nimbus DD2.0 */
1191     k->cores_mask = POWER9_CORE_MASK;
1192     k->core_pir = pnv_chip_core_pir_p9;
1193     k->intc_create = pnv_chip_power9_intc_create;
1194     k->isa_create = pnv_chip_power9_isa_create;
1195     k->dt_populate = pnv_chip_power9_dt_populate;
1196     k->pic_print_info = pnv_chip_power9_pic_print_info;
1197     dc->desc = "PowerNV Chip POWER9";
1198 
1199     device_class_set_parent_realize(dc, pnv_chip_power9_realize,
1200                                     &k->parent_realize);
1201 }
1202 
1203 static void pnv_chip_core_sanitize(PnvChip *chip, Error **errp)
1204 {
1205     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
1206     int cores_max;
1207 
1208     /*
1209      * No custom mask for this chip, let's use the default one from *
1210      * the chip class
1211      */
1212     if (!chip->cores_mask) {
1213         chip->cores_mask = pcc->cores_mask;
1214     }
1215 
1216     /* filter alien core ids ! some are reserved */
1217     if ((chip->cores_mask & pcc->cores_mask) != chip->cores_mask) {
1218         error_setg(errp, "warning: invalid core mask for chip Ox%"PRIx64" !",
1219                    chip->cores_mask);
1220         return;
1221     }
1222     chip->cores_mask &= pcc->cores_mask;
1223 
1224     /* now that we have a sane layout, let check the number of cores */
1225     cores_max = ctpop64(chip->cores_mask);
1226     if (chip->nr_cores > cores_max) {
1227         error_setg(errp, "warning: too many cores for chip ! Limit is %d",
1228                    cores_max);
1229         return;
1230     }
1231 }
1232 
1233 static void pnv_chip_core_realize(PnvChip *chip, Error **errp)
1234 {
1235     MachineState *ms = MACHINE(qdev_get_machine());
1236     Error *error = NULL;
1237     PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
1238     const char *typename = pnv_chip_core_typename(chip);
1239     size_t typesize = object_type_get_instance_size(typename);
1240     int i, core_hwid;
1241 
1242     if (!object_class_by_name(typename)) {
1243         error_setg(errp, "Unable to find PowerNV CPU Core '%s'", typename);
1244         return;
1245     }
1246 
1247     /* Cores */
1248     pnv_chip_core_sanitize(chip, &error);
1249     if (error) {
1250         error_propagate(errp, error);
1251         return;
1252     }
1253 
1254     chip->cores = g_malloc0(typesize * chip->nr_cores);
1255 
1256     for (i = 0, core_hwid = 0; (core_hwid < sizeof(chip->cores_mask) * 8)
1257              && (i < chip->nr_cores); core_hwid++) {
1258         char core_name[32];
1259         void *pnv_core = chip->cores + i * typesize;
1260         uint64_t xscom_core_base;
1261 
1262         if (!(chip->cores_mask & (1ull << core_hwid))) {
1263             continue;
1264         }
1265 
1266         snprintf(core_name, sizeof(core_name), "core[%d]", core_hwid);
1267         object_initialize_child(OBJECT(chip), core_name, pnv_core, typesize,
1268                                 typename, &error_fatal, NULL);
1269         object_property_set_int(OBJECT(pnv_core), ms->smp.threads, "nr-threads",
1270                                 &error_fatal);
1271         object_property_set_int(OBJECT(pnv_core), core_hwid,
1272                                 CPU_CORE_PROP_CORE_ID, &error_fatal);
1273         object_property_set_int(OBJECT(pnv_core),
1274                                 pcc->core_pir(chip, core_hwid),
1275                                 "pir", &error_fatal);
1276         object_property_add_const_link(OBJECT(pnv_core), "chip",
1277                                        OBJECT(chip), &error_fatal);
1278         object_property_set_bool(OBJECT(pnv_core), true, "realized",
1279                                  &error_fatal);
1280 
1281         /* Each core has an XSCOM MMIO region */
1282         if (!pnv_chip_is_power9(chip)) {
1283             xscom_core_base = PNV_XSCOM_EX_BASE(core_hwid);
1284         } else {
1285             xscom_core_base = PNV9_XSCOM_EC_BASE(core_hwid);
1286         }
1287 
1288         pnv_xscom_add_subregion(chip, xscom_core_base,
1289                                 &PNV_CORE(pnv_core)->xscom_regs);
1290         i++;
1291     }
1292 }
1293 
1294 static void pnv_chip_realize(DeviceState *dev, Error **errp)
1295 {
1296     PnvChip *chip = PNV_CHIP(dev);
1297     Error *error = NULL;
1298 
1299     /* Cores */
1300     pnv_chip_core_realize(chip, &error);
1301     if (error) {
1302         error_propagate(errp, error);
1303         return;
1304     }
1305 }
1306 
1307 static Property pnv_chip_properties[] = {
1308     DEFINE_PROP_UINT32("chip-id", PnvChip, chip_id, 0),
1309     DEFINE_PROP_UINT64("ram-start", PnvChip, ram_start, 0),
1310     DEFINE_PROP_UINT64("ram-size", PnvChip, ram_size, 0),
1311     DEFINE_PROP_UINT32("nr-cores", PnvChip, nr_cores, 1),
1312     DEFINE_PROP_UINT64("cores-mask", PnvChip, cores_mask, 0x0),
1313     DEFINE_PROP_END_OF_LIST(),
1314 };
1315 
1316 static void pnv_chip_class_init(ObjectClass *klass, void *data)
1317 {
1318     DeviceClass *dc = DEVICE_CLASS(klass);
1319 
1320     set_bit(DEVICE_CATEGORY_CPU, dc->categories);
1321     dc->realize = pnv_chip_realize;
1322     dc->props = pnv_chip_properties;
1323     dc->desc = "PowerNV Chip";
1324 }
1325 
1326 static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
1327 {
1328     PnvMachineState *pnv = PNV_MACHINE(xi);
1329     int i;
1330 
1331     for (i = 0; i < pnv->num_chips; i++) {
1332         Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
1333 
1334         if (ics_valid_irq(&chip8->psi.ics, irq)) {
1335             return &chip8->psi.ics;
1336         }
1337     }
1338     return NULL;
1339 }
1340 
1341 static void pnv_ics_resend(XICSFabric *xi)
1342 {
1343     PnvMachineState *pnv = PNV_MACHINE(xi);
1344     int i;
1345 
1346     for (i = 0; i < pnv->num_chips; i++) {
1347         Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
1348         ics_resend(&chip8->psi.ics);
1349     }
1350 }
1351 
1352 static ICPState *pnv_icp_get(XICSFabric *xi, int pir)
1353 {
1354     PowerPCCPU *cpu = ppc_get_vcpu_by_pir(pir);
1355 
1356     return cpu ? ICP(pnv_cpu_state(cpu)->intc) : NULL;
1357 }
1358 
1359 static void pnv_pic_print_info(InterruptStatsProvider *obj,
1360                                Monitor *mon)
1361 {
1362     PnvMachineState *pnv = PNV_MACHINE(obj);
1363     int i;
1364     CPUState *cs;
1365 
1366     CPU_FOREACH(cs) {
1367         PowerPCCPU *cpu = POWERPC_CPU(cs);
1368 
1369         if (pnv_chip_is_power9(pnv->chips[0])) {
1370             xive_tctx_pic_print_info(XIVE_TCTX(pnv_cpu_state(cpu)->intc), mon);
1371         } else {
1372             icp_pic_print_info(ICP(pnv_cpu_state(cpu)->intc), mon);
1373         }
1374     }
1375 
1376     for (i = 0; i < pnv->num_chips; i++) {
1377         PNV_CHIP_GET_CLASS(pnv->chips[i])->pic_print_info(pnv->chips[i], mon);
1378     }
1379 }
1380 
1381 static void pnv_get_num_chips(Object *obj, Visitor *v, const char *name,
1382                               void *opaque, Error **errp)
1383 {
1384     visit_type_uint32(v, name, &PNV_MACHINE(obj)->num_chips, errp);
1385 }
1386 
1387 static void pnv_set_num_chips(Object *obj, Visitor *v, const char *name,
1388                               void *opaque, Error **errp)
1389 {
1390     PnvMachineState *pnv = PNV_MACHINE(obj);
1391     uint32_t num_chips;
1392     Error *local_err = NULL;
1393 
1394     visit_type_uint32(v, name, &num_chips, &local_err);
1395     if (local_err) {
1396         error_propagate(errp, local_err);
1397         return;
1398     }
1399 
1400     /*
1401      * TODO: should we decide on how many chips we can create based
1402      * on #cores and Venice vs. Murano vs. Naples chip type etc...,
1403      */
1404     if (!is_power_of_2(num_chips) || num_chips > 4) {
1405         error_setg(errp, "invalid number of chips: '%d'", num_chips);
1406         return;
1407     }
1408 
1409     pnv->num_chips = num_chips;
1410 }
1411 
1412 static void pnv_machine_instance_init(Object *obj)
1413 {
1414     PnvMachineState *pnv = PNV_MACHINE(obj);
1415     pnv->num_chips = 1;
1416 }
1417 
1418 static void pnv_machine_class_props_init(ObjectClass *oc)
1419 {
1420     object_class_property_add(oc, "num-chips", "uint32",
1421                               pnv_get_num_chips, pnv_set_num_chips,
1422                               NULL, NULL, NULL);
1423     object_class_property_set_description(oc, "num-chips",
1424                               "Specifies the number of processor chips",
1425                               NULL);
1426 }
1427 
1428 static void pnv_machine_power8_class_init(ObjectClass *oc, void *data)
1429 {
1430     MachineClass *mc = MACHINE_CLASS(oc);
1431     XICSFabricClass *xic = XICS_FABRIC_CLASS(oc);
1432 
1433     mc->desc = "IBM PowerNV (Non-Virtualized) POWER8";
1434     mc->default_cpu_type = POWERPC_CPU_TYPE_NAME("power8_v2.0");
1435 
1436     xic->icp_get = pnv_icp_get;
1437     xic->ics_get = pnv_ics_get;
1438     xic->ics_resend = pnv_ics_resend;
1439 }
1440 
1441 static void pnv_machine_power9_class_init(ObjectClass *oc, void *data)
1442 {
1443     MachineClass *mc = MACHINE_CLASS(oc);
1444 
1445     mc->desc = "IBM PowerNV (Non-Virtualized) POWER9";
1446     mc->default_cpu_type = POWERPC_CPU_TYPE_NAME("power9_v2.0");
1447 
1448     mc->alias = "powernv";
1449 }
1450 
1451 static void pnv_machine_class_init(ObjectClass *oc, void *data)
1452 {
1453     MachineClass *mc = MACHINE_CLASS(oc);
1454     InterruptStatsProviderClass *ispc = INTERRUPT_STATS_PROVIDER_CLASS(oc);
1455 
1456     mc->desc = "IBM PowerNV (Non-Virtualized)";
1457     mc->init = pnv_init;
1458     mc->reset = pnv_reset;
1459     mc->max_cpus = MAX_CPUS;
1460     /* Pnv provides a AHCI device for storage */
1461     mc->block_default_type = IF_IDE;
1462     mc->no_parallel = 1;
1463     mc->default_boot_order = NULL;
1464     /*
1465      * RAM defaults to less than 2048 for 32-bit hosts, and large
1466      * enough to fit the maximum initrd size at it's load address
1467      */
1468     mc->default_ram_size = INITRD_LOAD_ADDR + INITRD_MAX_SIZE;
1469     ispc->print_info = pnv_pic_print_info;
1470 
1471     pnv_machine_class_props_init(oc);
1472 }
1473 
1474 #define DEFINE_PNV8_CHIP_TYPE(type, class_initfn) \
1475     {                                             \
1476         .name          = type,                    \
1477         .class_init    = class_initfn,            \
1478         .parent        = TYPE_PNV8_CHIP,          \
1479     }
1480 
1481 #define DEFINE_PNV9_CHIP_TYPE(type, class_initfn) \
1482     {                                             \
1483         .name          = type,                    \
1484         .class_init    = class_initfn,            \
1485         .parent        = TYPE_PNV9_CHIP,          \
1486     }
1487 
1488 static const TypeInfo types[] = {
1489     {
1490         .name          = MACHINE_TYPE_NAME("powernv9"),
1491         .parent        = TYPE_PNV_MACHINE,
1492         .class_init    = pnv_machine_power9_class_init,
1493     },
1494     {
1495         .name          = MACHINE_TYPE_NAME("powernv8"),
1496         .parent        = TYPE_PNV_MACHINE,
1497         .class_init    = pnv_machine_power8_class_init,
1498         .interfaces = (InterfaceInfo[]) {
1499             { TYPE_XICS_FABRIC },
1500             { },
1501         },
1502     },
1503     {
1504         .name          = TYPE_PNV_MACHINE,
1505         .parent        = TYPE_MACHINE,
1506         .abstract       = true,
1507         .instance_size = sizeof(PnvMachineState),
1508         .instance_init = pnv_machine_instance_init,
1509         .class_init    = pnv_machine_class_init,
1510         .interfaces = (InterfaceInfo[]) {
1511             { TYPE_INTERRUPT_STATS_PROVIDER },
1512             { },
1513         },
1514     },
1515     {
1516         .name          = TYPE_PNV_CHIP,
1517         .parent        = TYPE_SYS_BUS_DEVICE,
1518         .class_init    = pnv_chip_class_init,
1519         .instance_size = sizeof(PnvChip),
1520         .class_size    = sizeof(PnvChipClass),
1521         .abstract      = true,
1522     },
1523 
1524     /*
1525      * P9 chip and variants
1526      */
1527     {
1528         .name          = TYPE_PNV9_CHIP,
1529         .parent        = TYPE_PNV_CHIP,
1530         .instance_init = pnv_chip_power9_instance_init,
1531         .instance_size = sizeof(Pnv9Chip),
1532     },
1533     DEFINE_PNV9_CHIP_TYPE(TYPE_PNV_CHIP_POWER9, pnv_chip_power9_class_init),
1534 
1535     /*
1536      * P8 chip and variants
1537      */
1538     {
1539         .name          = TYPE_PNV8_CHIP,
1540         .parent        = TYPE_PNV_CHIP,
1541         .instance_init = pnv_chip_power8_instance_init,
1542         .instance_size = sizeof(Pnv8Chip),
1543     },
1544     DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8, pnv_chip_power8_class_init),
1545     DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8E, pnv_chip_power8e_class_init),
1546     DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8NVL,
1547                           pnv_chip_power8nvl_class_init),
1548 };
1549 
1550 DEFINE_TYPES(types)
1551