test-bdrv-drain: AIO_WAIT_WHILE() in job .commit/.abort
[qemu/ar7.git] / hw / ppc / pnv.c
blob346f5e7aedb5b282dd1253b502196f4c1a911ce8
1 /*
2 * QEMU PowerPC PowerNV machine model
4 * Copyright (c) 2016, IBM Corporation.
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.
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.
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/>.
20 #include "qemu/osdep.h"
21 #include "qemu/units.h"
22 #include "qapi/error.h"
23 #include "sysemu/sysemu.h"
24 #include "sysemu/numa.h"
25 #include "sysemu/cpus.h"
26 #include "hw/hw.h"
27 #include "target/ppc/cpu.h"
28 #include "qemu/log.h"
29 #include "hw/ppc/fdt.h"
30 #include "hw/ppc/ppc.h"
31 #include "hw/ppc/pnv.h"
32 #include "hw/ppc/pnv_core.h"
33 #include "hw/loader.h"
34 #include "exec/address-spaces.h"
35 #include "qapi/visitor.h"
36 #include "monitor/monitor.h"
37 #include "hw/intc/intc.h"
38 #include "hw/ipmi/ipmi.h"
39 #include "target/ppc/mmu-hash64.h"
41 #include "hw/ppc/xics.h"
42 #include "hw/ppc/pnv_xscom.h"
44 #include "hw/isa/isa.h"
45 #include "hw/char/serial.h"
46 #include "hw/timer/mc146818rtc.h"
48 #include <libfdt.h>
50 #define FDT_MAX_SIZE 0x00100000
52 #define FW_FILE_NAME "skiboot.lid"
53 #define FW_LOAD_ADDR 0x0
54 #define FW_MAX_SIZE 0x00400000
56 #define KERNEL_LOAD_ADDR 0x20000000
57 #define INITRD_LOAD_ADDR 0x60000000
59 static const char *pnv_chip_core_typename(const PnvChip *o)
61 const char *chip_type = object_class_get_name(object_get_class(OBJECT(o)));
62 int len = strlen(chip_type) - strlen(PNV_CHIP_TYPE_SUFFIX);
63 char *s = g_strdup_printf(PNV_CORE_TYPE_NAME("%.*s"), len, chip_type);
64 const char *core_type = object_class_get_name(object_class_by_name(s));
65 g_free(s);
66 return core_type;
70 * On Power Systems E880 (POWER8), the max cpus (threads) should be :
71 * 4 * 4 sockets * 12 cores * 8 threads = 1536
72 * Let's make it 2^11
74 #define MAX_CPUS 2048
77 * Memory nodes are created by hostboot, one for each range of memory
78 * that has a different "affinity". In practice, it means one range
79 * per chip.
81 static void pnv_dt_memory(void *fdt, int chip_id, hwaddr start, hwaddr size)
83 char *mem_name;
84 uint64_t mem_reg_property[2];
85 int off;
87 mem_reg_property[0] = cpu_to_be64(start);
88 mem_reg_property[1] = cpu_to_be64(size);
90 mem_name = g_strdup_printf("memory@%"HWADDR_PRIx, start);
91 off = fdt_add_subnode(fdt, 0, mem_name);
92 g_free(mem_name);
94 _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
95 _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
96 sizeof(mem_reg_property))));
97 _FDT((fdt_setprop_cell(fdt, off, "ibm,chip-id", chip_id)));
100 static int get_cpus_node(void *fdt)
102 int cpus_offset = fdt_path_offset(fdt, "/cpus");
104 if (cpus_offset < 0) {
105 cpus_offset = fdt_add_subnode(fdt, 0, "cpus");
106 if (cpus_offset) {
107 _FDT((fdt_setprop_cell(fdt, cpus_offset, "#address-cells", 0x1)));
108 _FDT((fdt_setprop_cell(fdt, cpus_offset, "#size-cells", 0x0)));
111 _FDT(cpus_offset);
112 return cpus_offset;
116 * The PowerNV cores (and threads) need to use real HW ids and not an
117 * incremental index like it has been done on other platforms. This HW
118 * id is stored in the CPU PIR, it is used to create cpu nodes in the
119 * device tree, used in XSCOM to address cores and in interrupt
120 * servers.
122 static void pnv_dt_core(PnvChip *chip, PnvCore *pc, void *fdt)
124 PowerPCCPU *cpu = pc->threads[0];
125 CPUState *cs = CPU(cpu);
126 DeviceClass *dc = DEVICE_GET_CLASS(cs);
127 int smt_threads = CPU_CORE(pc)->nr_threads;
128 CPUPPCState *env = &cpu->env;
129 PowerPCCPUClass *pcc = POWERPC_CPU_GET_CLASS(cs);
130 uint32_t servers_prop[smt_threads];
131 int i;
132 uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
133 0xffffffff, 0xffffffff};
134 uint32_t tbfreq = PNV_TIMEBASE_FREQ;
135 uint32_t cpufreq = 1000000000;
136 uint32_t page_sizes_prop[64];
137 size_t page_sizes_prop_size;
138 const uint8_t pa_features[] = { 24, 0,
139 0xf6, 0x3f, 0xc7, 0xc0, 0x80, 0xf0,
140 0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
141 0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
142 0x80, 0x00, 0x80, 0x00, 0x80, 0x00 };
143 int offset;
144 char *nodename;
145 int cpus_offset = get_cpus_node(fdt);
147 nodename = g_strdup_printf("%s@%x", dc->fw_name, pc->pir);
148 offset = fdt_add_subnode(fdt, cpus_offset, nodename);
149 _FDT(offset);
150 g_free(nodename);
152 _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id", chip->chip_id)));
154 _FDT((fdt_setprop_cell(fdt, offset, "reg", pc->pir)));
155 _FDT((fdt_setprop_cell(fdt, offset, "ibm,pir", pc->pir)));
156 _FDT((fdt_setprop_string(fdt, offset, "device_type", "cpu")));
158 _FDT((fdt_setprop_cell(fdt, offset, "cpu-version", env->spr[SPR_PVR])));
159 _FDT((fdt_setprop_cell(fdt, offset, "d-cache-block-size",
160 env->dcache_line_size)));
161 _FDT((fdt_setprop_cell(fdt, offset, "d-cache-line-size",
162 env->dcache_line_size)));
163 _FDT((fdt_setprop_cell(fdt, offset, "i-cache-block-size",
164 env->icache_line_size)));
165 _FDT((fdt_setprop_cell(fdt, offset, "i-cache-line-size",
166 env->icache_line_size)));
168 if (pcc->l1_dcache_size) {
169 _FDT((fdt_setprop_cell(fdt, offset, "d-cache-size",
170 pcc->l1_dcache_size)));
171 } else {
172 warn_report("Unknown L1 dcache size for cpu");
174 if (pcc->l1_icache_size) {
175 _FDT((fdt_setprop_cell(fdt, offset, "i-cache-size",
176 pcc->l1_icache_size)));
177 } else {
178 warn_report("Unknown L1 icache size for cpu");
181 _FDT((fdt_setprop_cell(fdt, offset, "timebase-frequency", tbfreq)));
182 _FDT((fdt_setprop_cell(fdt, offset, "clock-frequency", cpufreq)));
183 _FDT((fdt_setprop_cell(fdt, offset, "ibm,slb-size", cpu->hash64_opts->slb_size)));
184 _FDT((fdt_setprop_string(fdt, offset, "status", "okay")));
185 _FDT((fdt_setprop(fdt, offset, "64-bit", NULL, 0)));
187 if (env->spr_cb[SPR_PURR].oea_read) {
188 _FDT((fdt_setprop(fdt, offset, "ibm,purr", NULL, 0)));
191 if (ppc_hash64_has(cpu, PPC_HASH64_1TSEG)) {
192 _FDT((fdt_setprop(fdt, offset, "ibm,processor-segment-sizes",
193 segs, sizeof(segs))));
196 /* Advertise VMX/VSX (vector extensions) if available
197 * 0 / no property == no vector extensions
198 * 1 == VMX / Altivec available
199 * 2 == VSX available */
200 if (env->insns_flags & PPC_ALTIVEC) {
201 uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
203 _FDT((fdt_setprop_cell(fdt, offset, "ibm,vmx", vmx)));
206 /* Advertise DFP (Decimal Floating Point) if available
207 * 0 / no property == no DFP
208 * 1 == DFP available */
209 if (env->insns_flags2 & PPC2_DFP) {
210 _FDT((fdt_setprop_cell(fdt, offset, "ibm,dfp", 1)));
213 page_sizes_prop_size = ppc_create_page_sizes_prop(cpu, page_sizes_prop,
214 sizeof(page_sizes_prop));
215 if (page_sizes_prop_size) {
216 _FDT((fdt_setprop(fdt, offset, "ibm,segment-page-sizes",
217 page_sizes_prop, page_sizes_prop_size)));
220 _FDT((fdt_setprop(fdt, offset, "ibm,pa-features",
221 pa_features, sizeof(pa_features))));
223 /* Build interrupt servers properties */
224 for (i = 0; i < smt_threads; i++) {
225 servers_prop[i] = cpu_to_be32(pc->pir + i);
227 _FDT((fdt_setprop(fdt, offset, "ibm,ppc-interrupt-server#s",
228 servers_prop, sizeof(servers_prop))));
231 static void pnv_dt_icp(PnvChip *chip, void *fdt, uint32_t pir,
232 uint32_t nr_threads)
234 uint64_t addr = PNV_ICP_BASE(chip) | (pir << 12);
235 char *name;
236 const char compat[] = "IBM,power8-icp\0IBM,ppc-xicp";
237 uint32_t irange[2], i, rsize;
238 uint64_t *reg;
239 int offset;
241 irange[0] = cpu_to_be32(pir);
242 irange[1] = cpu_to_be32(nr_threads);
244 rsize = sizeof(uint64_t) * 2 * nr_threads;
245 reg = g_malloc(rsize);
246 for (i = 0; i < nr_threads; i++) {
247 reg[i * 2] = cpu_to_be64(addr | ((pir + i) * 0x1000));
248 reg[i * 2 + 1] = cpu_to_be64(0x1000);
251 name = g_strdup_printf("interrupt-controller@%"PRIX64, addr);
252 offset = fdt_add_subnode(fdt, 0, name);
253 _FDT(offset);
254 g_free(name);
256 _FDT((fdt_setprop(fdt, offset, "compatible", compat, sizeof(compat))));
257 _FDT((fdt_setprop(fdt, offset, "reg", reg, rsize)));
258 _FDT((fdt_setprop_string(fdt, offset, "device_type",
259 "PowerPC-External-Interrupt-Presentation")));
260 _FDT((fdt_setprop(fdt, offset, "interrupt-controller", NULL, 0)));
261 _FDT((fdt_setprop(fdt, offset, "ibm,interrupt-server-ranges",
262 irange, sizeof(irange))));
263 _FDT((fdt_setprop_cell(fdt, offset, "#interrupt-cells", 1)));
264 _FDT((fdt_setprop_cell(fdt, offset, "#address-cells", 0)));
265 g_free(reg);
268 static void pnv_dt_chip(PnvChip *chip, void *fdt)
270 const char *typename = pnv_chip_core_typename(chip);
271 size_t typesize = object_type_get_instance_size(typename);
272 int i;
274 pnv_dt_xscom(chip, fdt, 0);
276 for (i = 0; i < chip->nr_cores; i++) {
277 PnvCore *pnv_core = PNV_CORE(chip->cores + i * typesize);
279 pnv_dt_core(chip, pnv_core, fdt);
281 /* Interrupt Control Presenters (ICP). One per core. */
282 pnv_dt_icp(chip, fdt, pnv_core->pir, CPU_CORE(pnv_core)->nr_threads);
285 if (chip->ram_size) {
286 pnv_dt_memory(fdt, chip->chip_id, chip->ram_start, chip->ram_size);
290 static void pnv_dt_rtc(ISADevice *d, void *fdt, int lpc_off)
292 uint32_t io_base = d->ioport_id;
293 uint32_t io_regs[] = {
294 cpu_to_be32(1),
295 cpu_to_be32(io_base),
296 cpu_to_be32(2)
298 char *name;
299 int node;
301 name = g_strdup_printf("%s@i%x", qdev_fw_name(DEVICE(d)), io_base);
302 node = fdt_add_subnode(fdt, lpc_off, name);
303 _FDT(node);
304 g_free(name);
306 _FDT((fdt_setprop(fdt, node, "reg", io_regs, sizeof(io_regs))));
307 _FDT((fdt_setprop_string(fdt, node, "compatible", "pnpPNP,b00")));
310 static void pnv_dt_serial(ISADevice *d, void *fdt, int lpc_off)
312 const char compatible[] = "ns16550\0pnpPNP,501";
313 uint32_t io_base = d->ioport_id;
314 uint32_t io_regs[] = {
315 cpu_to_be32(1),
316 cpu_to_be32(io_base),
317 cpu_to_be32(8)
319 char *name;
320 int node;
322 name = g_strdup_printf("%s@i%x", qdev_fw_name(DEVICE(d)), io_base);
323 node = fdt_add_subnode(fdt, lpc_off, name);
324 _FDT(node);
325 g_free(name);
327 _FDT((fdt_setprop(fdt, node, "reg", io_regs, sizeof(io_regs))));
328 _FDT((fdt_setprop(fdt, node, "compatible", compatible,
329 sizeof(compatible))));
331 _FDT((fdt_setprop_cell(fdt, node, "clock-frequency", 1843200)));
332 _FDT((fdt_setprop_cell(fdt, node, "current-speed", 115200)));
333 _FDT((fdt_setprop_cell(fdt, node, "interrupts", d->isairq[0])));
334 _FDT((fdt_setprop_cell(fdt, node, "interrupt-parent",
335 fdt_get_phandle(fdt, lpc_off))));
337 /* This is needed by Linux */
338 _FDT((fdt_setprop_string(fdt, node, "device_type", "serial")));
341 static void pnv_dt_ipmi_bt(ISADevice *d, void *fdt, int lpc_off)
343 const char compatible[] = "bt\0ipmi-bt";
344 uint32_t io_base;
345 uint32_t io_regs[] = {
346 cpu_to_be32(1),
347 0, /* 'io_base' retrieved from the 'ioport' property of 'isa-ipmi-bt' */
348 cpu_to_be32(3)
350 uint32_t irq;
351 char *name;
352 int node;
354 io_base = object_property_get_int(OBJECT(d), "ioport", &error_fatal);
355 io_regs[1] = cpu_to_be32(io_base);
357 irq = object_property_get_int(OBJECT(d), "irq", &error_fatal);
359 name = g_strdup_printf("%s@i%x", qdev_fw_name(DEVICE(d)), io_base);
360 node = fdt_add_subnode(fdt, lpc_off, name);
361 _FDT(node);
362 g_free(name);
364 _FDT((fdt_setprop(fdt, node, "reg", io_regs, sizeof(io_regs))));
365 _FDT((fdt_setprop(fdt, node, "compatible", compatible,
366 sizeof(compatible))));
368 /* Mark it as reserved to avoid Linux trying to claim it */
369 _FDT((fdt_setprop_string(fdt, node, "status", "reserved")));
370 _FDT((fdt_setprop_cell(fdt, node, "interrupts", irq)));
371 _FDT((fdt_setprop_cell(fdt, node, "interrupt-parent",
372 fdt_get_phandle(fdt, lpc_off))));
375 typedef struct ForeachPopulateArgs {
376 void *fdt;
377 int offset;
378 } ForeachPopulateArgs;
380 static int pnv_dt_isa_device(DeviceState *dev, void *opaque)
382 ForeachPopulateArgs *args = opaque;
383 ISADevice *d = ISA_DEVICE(dev);
385 if (object_dynamic_cast(OBJECT(dev), TYPE_MC146818_RTC)) {
386 pnv_dt_rtc(d, args->fdt, args->offset);
387 } else if (object_dynamic_cast(OBJECT(dev), TYPE_ISA_SERIAL)) {
388 pnv_dt_serial(d, args->fdt, args->offset);
389 } else if (object_dynamic_cast(OBJECT(dev), "isa-ipmi-bt")) {
390 pnv_dt_ipmi_bt(d, args->fdt, args->offset);
391 } else {
392 error_report("unknown isa device %s@i%x", qdev_fw_name(dev),
393 d->ioport_id);
396 return 0;
399 static int pnv_chip_isa_offset(PnvChip *chip, void *fdt)
401 char *name;
402 int offset;
404 name = g_strdup_printf("/xscom@%" PRIx64 "/isa@%x",
405 (uint64_t) PNV_XSCOM_BASE(chip), PNV_XSCOM_LPC_BASE);
406 offset = fdt_path_offset(fdt, name);
407 g_free(name);
408 return offset;
411 /* The default LPC bus of a multichip system is on chip 0. It's
412 * recognized by the firmware (skiboot) using a "primary" property.
414 static void pnv_dt_isa(PnvMachineState *pnv, void *fdt)
416 int isa_offset = pnv_chip_isa_offset(pnv->chips[0], fdt);
417 ForeachPopulateArgs args = {
418 .fdt = fdt,
419 .offset = isa_offset,
422 _FDT((fdt_setprop(fdt, isa_offset, "primary", NULL, 0)));
424 /* ISA devices are not necessarily parented to the ISA bus so we
425 * can not use object_child_foreach() */
426 qbus_walk_children(BUS(pnv->isa_bus), pnv_dt_isa_device, NULL, NULL, NULL,
427 &args);
430 static void *pnv_dt_create(MachineState *machine)
432 const char plat_compat[] = "qemu,powernv\0ibm,powernv";
433 PnvMachineState *pnv = PNV_MACHINE(machine);
434 void *fdt;
435 char *buf;
436 int off;
437 int i;
439 fdt = g_malloc0(FDT_MAX_SIZE);
440 _FDT((fdt_create_empty_tree(fdt, FDT_MAX_SIZE)));
442 /* Root node */
443 _FDT((fdt_setprop_cell(fdt, 0, "#address-cells", 0x2)));
444 _FDT((fdt_setprop_cell(fdt, 0, "#size-cells", 0x2)));
445 _FDT((fdt_setprop_string(fdt, 0, "model",
446 "IBM PowerNV (emulated by qemu)")));
447 _FDT((fdt_setprop(fdt, 0, "compatible", plat_compat,
448 sizeof(plat_compat))));
450 buf = qemu_uuid_unparse_strdup(&qemu_uuid);
451 _FDT((fdt_setprop_string(fdt, 0, "vm,uuid", buf)));
452 if (qemu_uuid_set) {
453 _FDT((fdt_property_string(fdt, "system-id", buf)));
455 g_free(buf);
457 off = fdt_add_subnode(fdt, 0, "chosen");
458 if (machine->kernel_cmdline) {
459 _FDT((fdt_setprop_string(fdt, off, "bootargs",
460 machine->kernel_cmdline)));
463 if (pnv->initrd_size) {
464 uint32_t start_prop = cpu_to_be32(pnv->initrd_base);
465 uint32_t end_prop = cpu_to_be32(pnv->initrd_base + pnv->initrd_size);
467 _FDT((fdt_setprop(fdt, off, "linux,initrd-start",
468 &start_prop, sizeof(start_prop))));
469 _FDT((fdt_setprop(fdt, off, "linux,initrd-end",
470 &end_prop, sizeof(end_prop))));
473 /* Populate device tree for each chip */
474 for (i = 0; i < pnv->num_chips; i++) {
475 pnv_dt_chip(pnv->chips[i], fdt);
478 /* Populate ISA devices on chip 0 */
479 pnv_dt_isa(pnv, fdt);
481 if (pnv->bmc) {
482 pnv_dt_bmc_sensors(pnv->bmc, fdt);
485 return fdt;
488 static void pnv_powerdown_notify(Notifier *n, void *opaque)
490 PnvMachineState *pnv = PNV_MACHINE(qdev_get_machine());
492 if (pnv->bmc) {
493 pnv_bmc_powerdown(pnv->bmc);
497 static void pnv_reset(void)
499 MachineState *machine = MACHINE(qdev_get_machine());
500 PnvMachineState *pnv = PNV_MACHINE(machine);
501 void *fdt;
502 Object *obj;
504 qemu_devices_reset();
506 /* OpenPOWER systems have a BMC, which can be defined on the
507 * command line with:
509 * -device ipmi-bmc-sim,id=bmc0
511 * This is the internal simulator but it could also be an external
512 * BMC.
514 obj = object_resolve_path_type("", "ipmi-bmc-sim", NULL);
515 if (obj) {
516 pnv->bmc = IPMI_BMC(obj);
519 fdt = pnv_dt_create(machine);
521 /* Pack resulting tree */
522 _FDT((fdt_pack(fdt)));
524 cpu_physical_memory_write(PNV_FDT_ADDR, fdt, fdt_totalsize(fdt));
527 static ISABus *pnv_chip_power8_isa_create(PnvChip *chip, Error **errp)
529 Pnv8Chip *chip8 = PNV8_CHIP(chip);
530 return pnv_lpc_isa_create(&chip8->lpc, true, errp);
533 static ISABus *pnv_chip_power8nvl_isa_create(PnvChip *chip, Error **errp)
535 Pnv8Chip *chip8 = PNV8_CHIP(chip);
536 return pnv_lpc_isa_create(&chip8->lpc, false, errp);
539 static ISABus *pnv_chip_power9_isa_create(PnvChip *chip, Error **errp)
541 return NULL;
544 static ISABus *pnv_isa_create(PnvChip *chip, Error **errp)
546 return PNV_CHIP_GET_CLASS(chip)->isa_create(chip, errp);
549 static void pnv_init(MachineState *machine)
551 PnvMachineState *pnv = PNV_MACHINE(machine);
552 MemoryRegion *ram;
553 char *fw_filename;
554 long fw_size;
555 int i;
556 char *chip_typename;
558 /* allocate RAM */
559 if (machine->ram_size < (1 * GiB)) {
560 warn_report("skiboot may not work with < 1GB of RAM");
563 ram = g_new(MemoryRegion, 1);
564 memory_region_allocate_system_memory(ram, NULL, "pnv.ram",
565 machine->ram_size);
566 memory_region_add_subregion(get_system_memory(), 0, ram);
568 /* load skiboot firmware */
569 if (bios_name == NULL) {
570 bios_name = FW_FILE_NAME;
573 fw_filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
574 if (!fw_filename) {
575 error_report("Could not find OPAL firmware '%s'", bios_name);
576 exit(1);
579 fw_size = load_image_targphys(fw_filename, FW_LOAD_ADDR, FW_MAX_SIZE);
580 if (fw_size < 0) {
581 error_report("Could not load OPAL firmware '%s'", fw_filename);
582 exit(1);
584 g_free(fw_filename);
586 /* load kernel */
587 if (machine->kernel_filename) {
588 long kernel_size;
590 kernel_size = load_image_targphys(machine->kernel_filename,
591 KERNEL_LOAD_ADDR, 0x2000000);
592 if (kernel_size < 0) {
593 error_report("Could not load kernel '%s'",
594 machine->kernel_filename);
595 exit(1);
599 /* load initrd */
600 if (machine->initrd_filename) {
601 pnv->initrd_base = INITRD_LOAD_ADDR;
602 pnv->initrd_size = load_image_targphys(machine->initrd_filename,
603 pnv->initrd_base, 0x10000000); /* 128MB max */
604 if (pnv->initrd_size < 0) {
605 error_report("Could not load initial ram disk '%s'",
606 machine->initrd_filename);
607 exit(1);
611 /* Create the processor chips */
612 i = strlen(machine->cpu_type) - strlen(POWERPC_CPU_TYPE_SUFFIX);
613 chip_typename = g_strdup_printf(PNV_CHIP_TYPE_NAME("%.*s"),
614 i, machine->cpu_type);
615 if (!object_class_by_name(chip_typename)) {
616 error_report("invalid CPU model '%.*s' for %s machine",
617 i, machine->cpu_type, MACHINE_GET_CLASS(machine)->name);
618 exit(1);
621 pnv->chips = g_new0(PnvChip *, pnv->num_chips);
622 for (i = 0; i < pnv->num_chips; i++) {
623 char chip_name[32];
624 Object *chip = object_new(chip_typename);
626 pnv->chips[i] = PNV_CHIP(chip);
628 /* TODO: put all the memory in one node on chip 0 until we find a
629 * way to specify different ranges for each chip
631 if (i == 0) {
632 object_property_set_int(chip, machine->ram_size, "ram-size",
633 &error_fatal);
636 snprintf(chip_name, sizeof(chip_name), "chip[%d]", PNV_CHIP_HWID(i));
637 object_property_add_child(OBJECT(pnv), chip_name, chip, &error_fatal);
638 object_property_set_int(chip, PNV_CHIP_HWID(i), "chip-id",
639 &error_fatal);
640 object_property_set_int(chip, smp_cores, "nr-cores", &error_fatal);
641 object_property_set_bool(chip, true, "realized", &error_fatal);
643 g_free(chip_typename);
645 /* Instantiate ISA bus on chip 0 */
646 pnv->isa_bus = pnv_isa_create(pnv->chips[0], &error_fatal);
648 /* Create serial port */
649 serial_hds_isa_init(pnv->isa_bus, 0, MAX_ISA_SERIAL_PORTS);
651 /* Create an RTC ISA device too */
652 mc146818_rtc_init(pnv->isa_bus, 2000, NULL);
654 /* OpenPOWER systems use a IPMI SEL Event message to notify the
655 * host to powerdown */
656 pnv->powerdown_notifier.notify = pnv_powerdown_notify;
657 qemu_register_powerdown_notifier(&pnv->powerdown_notifier);
661 * 0:21 Reserved - Read as zeros
662 * 22:24 Chip ID
663 * 25:28 Core number
664 * 29:31 Thread ID
666 static uint32_t pnv_chip_core_pir_p8(PnvChip *chip, uint32_t core_id)
668 return (chip->chip_id << 7) | (core_id << 3);
671 static Object *pnv_chip_power8_intc_create(PnvChip *chip, Object *child,
672 Error **errp)
674 return icp_create(child, TYPE_PNV_ICP, XICS_FABRIC(qdev_get_machine()),
675 errp);
679 * 0:48 Reserved - Read as zeroes
680 * 49:52 Node ID
681 * 53:55 Chip ID
682 * 56 Reserved - Read as zero
683 * 57:61 Core number
684 * 62:63 Thread ID
686 * We only care about the lower bits. uint32_t is fine for the moment.
688 static uint32_t pnv_chip_core_pir_p9(PnvChip *chip, uint32_t core_id)
690 return (chip->chip_id << 8) | (core_id << 2);
693 static Object *pnv_chip_power9_intc_create(PnvChip *chip, Object *child,
694 Error **errp)
696 return NULL;
699 /* Allowed core identifiers on a POWER8 Processor Chip :
701 * <EX0 reserved>
702 * EX1 - Venice only
703 * EX2 - Venice only
704 * EX3 - Venice only
705 * EX4
706 * EX5
707 * EX6
708 * <EX7,8 reserved> <reserved>
709 * EX9 - Venice only
710 * EX10 - Venice only
711 * EX11 - Venice only
712 * EX12
713 * EX13
714 * EX14
715 * <EX15 reserved>
717 #define POWER8E_CORE_MASK (0x7070ull)
718 #define POWER8_CORE_MASK (0x7e7eull)
721 * POWER9 has 24 cores, ids starting at 0x0
723 #define POWER9_CORE_MASK (0xffffffffffffffull)
725 static void pnv_chip_power8_instance_init(Object *obj)
727 Pnv8Chip *chip8 = PNV8_CHIP(obj);
729 object_initialize(&chip8->psi, sizeof(chip8->psi), TYPE_PNV_PSI);
730 object_property_add_child(obj, "psi", OBJECT(&chip8->psi), NULL);
731 object_property_add_const_link(OBJECT(&chip8->psi), "xics",
732 OBJECT(qdev_get_machine()), &error_abort);
734 object_initialize(&chip8->lpc, sizeof(chip8->lpc), TYPE_PNV_LPC);
735 object_property_add_child(obj, "lpc", OBJECT(&chip8->lpc), NULL);
736 object_property_add_const_link(OBJECT(&chip8->lpc), "psi",
737 OBJECT(&chip8->psi), &error_abort);
739 object_initialize(&chip8->occ, sizeof(chip8->occ), TYPE_PNV_OCC);
740 object_property_add_child(obj, "occ", OBJECT(&chip8->occ), NULL);
741 object_property_add_const_link(OBJECT(&chip8->occ), "psi",
742 OBJECT(&chip8->psi), &error_abort);
745 static void pnv_chip_icp_realize(Pnv8Chip *chip8, Error **errp)
747 PnvChip *chip = PNV_CHIP(chip8);
748 PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
749 const char *typename = pnv_chip_core_typename(chip);
750 size_t typesize = object_type_get_instance_size(typename);
751 int i, j;
752 char *name;
753 XICSFabric *xi = XICS_FABRIC(qdev_get_machine());
755 name = g_strdup_printf("icp-%x", chip->chip_id);
756 memory_region_init(&chip8->icp_mmio, OBJECT(chip), name, PNV_ICP_SIZE);
757 sysbus_init_mmio(SYS_BUS_DEVICE(chip), &chip8->icp_mmio);
758 g_free(name);
760 sysbus_mmio_map(SYS_BUS_DEVICE(chip), 1, PNV_ICP_BASE(chip));
762 /* Map the ICP registers for each thread */
763 for (i = 0; i < chip->nr_cores; i++) {
764 PnvCore *pnv_core = PNV_CORE(chip->cores + i * typesize);
765 int core_hwid = CPU_CORE(pnv_core)->core_id;
767 for (j = 0; j < CPU_CORE(pnv_core)->nr_threads; j++) {
768 uint32_t pir = pcc->core_pir(chip, core_hwid) + j;
769 PnvICPState *icp = PNV_ICP(xics_icp_get(xi, pir));
771 memory_region_add_subregion(&chip8->icp_mmio, pir << 12,
772 &icp->mmio);
777 static void pnv_chip_power8_realize(DeviceState *dev, Error **errp)
779 PnvChipClass *pcc = PNV_CHIP_GET_CLASS(dev);
780 PnvChip *chip = PNV_CHIP(dev);
781 Pnv8Chip *chip8 = PNV8_CHIP(dev);
782 Error *local_err = NULL;
784 pcc->parent_realize(dev, &local_err);
785 if (local_err) {
786 error_propagate(errp, local_err);
787 return;
790 /* Processor Service Interface (PSI) Host Bridge */
791 object_property_set_int(OBJECT(&chip8->psi), PNV_PSIHB_BASE(chip),
792 "bar", &error_fatal);
793 object_property_set_bool(OBJECT(&chip8->psi), true, "realized", &local_err);
794 if (local_err) {
795 error_propagate(errp, local_err);
796 return;
798 pnv_xscom_add_subregion(chip, PNV_XSCOM_PSIHB_BASE, &chip8->psi.xscom_regs);
800 /* Create LPC controller */
801 object_property_set_bool(OBJECT(&chip8->lpc), true, "realized",
802 &error_fatal);
803 pnv_xscom_add_subregion(chip, PNV_XSCOM_LPC_BASE, &chip8->lpc.xscom_regs);
805 /* Interrupt Management Area. This is the memory region holding
806 * all the Interrupt Control Presenter (ICP) registers */
807 pnv_chip_icp_realize(chip8, &local_err);
808 if (local_err) {
809 error_propagate(errp, local_err);
810 return;
813 /* Create the simplified OCC model */
814 object_property_set_bool(OBJECT(&chip8->occ), true, "realized", &local_err);
815 if (local_err) {
816 error_propagate(errp, local_err);
817 return;
819 pnv_xscom_add_subregion(chip, PNV_XSCOM_OCC_BASE, &chip8->occ.xscom_regs);
822 static void pnv_chip_power8e_class_init(ObjectClass *klass, void *data)
824 DeviceClass *dc = DEVICE_CLASS(klass);
825 PnvChipClass *k = PNV_CHIP_CLASS(klass);
827 k->chip_type = PNV_CHIP_POWER8E;
828 k->chip_cfam_id = 0x221ef04980000000ull; /* P8 Murano DD2.1 */
829 k->cores_mask = POWER8E_CORE_MASK;
830 k->core_pir = pnv_chip_core_pir_p8;
831 k->intc_create = pnv_chip_power8_intc_create;
832 k->isa_create = pnv_chip_power8_isa_create;
833 k->xscom_base = 0x003fc0000000000ull;
834 dc->desc = "PowerNV Chip POWER8E";
836 device_class_set_parent_realize(dc, pnv_chip_power8_realize,
837 &k->parent_realize);
840 static void pnv_chip_power8_class_init(ObjectClass *klass, void *data)
842 DeviceClass *dc = DEVICE_CLASS(klass);
843 PnvChipClass *k = PNV_CHIP_CLASS(klass);
845 k->chip_type = PNV_CHIP_POWER8;
846 k->chip_cfam_id = 0x220ea04980000000ull; /* P8 Venice DD2.0 */
847 k->cores_mask = POWER8_CORE_MASK;
848 k->core_pir = pnv_chip_core_pir_p8;
849 k->intc_create = pnv_chip_power8_intc_create;
850 k->isa_create = pnv_chip_power8_isa_create;
851 k->xscom_base = 0x003fc0000000000ull;
852 dc->desc = "PowerNV Chip POWER8";
854 device_class_set_parent_realize(dc, pnv_chip_power8_realize,
855 &k->parent_realize);
858 static void pnv_chip_power8nvl_class_init(ObjectClass *klass, void *data)
860 DeviceClass *dc = DEVICE_CLASS(klass);
861 PnvChipClass *k = PNV_CHIP_CLASS(klass);
863 k->chip_type = PNV_CHIP_POWER8NVL;
864 k->chip_cfam_id = 0x120d304980000000ull; /* P8 Naples DD1.0 */
865 k->cores_mask = POWER8_CORE_MASK;
866 k->core_pir = pnv_chip_core_pir_p8;
867 k->intc_create = pnv_chip_power8_intc_create;
868 k->isa_create = pnv_chip_power8nvl_isa_create;
869 k->xscom_base = 0x003fc0000000000ull;
870 dc->desc = "PowerNV Chip POWER8NVL";
872 device_class_set_parent_realize(dc, pnv_chip_power8_realize,
873 &k->parent_realize);
876 static void pnv_chip_power9_instance_init(Object *obj)
880 static void pnv_chip_power9_realize(DeviceState *dev, Error **errp)
882 PnvChipClass *pcc = PNV_CHIP_GET_CLASS(dev);
883 Error *local_err = NULL;
885 pcc->parent_realize(dev, &local_err);
886 if (local_err) {
887 error_propagate(errp, local_err);
888 return;
892 static void pnv_chip_power9_class_init(ObjectClass *klass, void *data)
894 DeviceClass *dc = DEVICE_CLASS(klass);
895 PnvChipClass *k = PNV_CHIP_CLASS(klass);
897 k->chip_type = PNV_CHIP_POWER9;
898 k->chip_cfam_id = 0x220d104900008000ull; /* P9 Nimbus DD2.0 */
899 k->cores_mask = POWER9_CORE_MASK;
900 k->core_pir = pnv_chip_core_pir_p9;
901 k->intc_create = pnv_chip_power9_intc_create;
902 k->isa_create = pnv_chip_power9_isa_create;
903 k->xscom_base = 0x00603fc00000000ull;
904 dc->desc = "PowerNV Chip POWER9";
906 device_class_set_parent_realize(dc, pnv_chip_power9_realize,
907 &k->parent_realize);
910 static void pnv_chip_core_sanitize(PnvChip *chip, Error **errp)
912 PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
913 int cores_max;
916 * No custom mask for this chip, let's use the default one from *
917 * the chip class
919 if (!chip->cores_mask) {
920 chip->cores_mask = pcc->cores_mask;
923 /* filter alien core ids ! some are reserved */
924 if ((chip->cores_mask & pcc->cores_mask) != chip->cores_mask) {
925 error_setg(errp, "warning: invalid core mask for chip Ox%"PRIx64" !",
926 chip->cores_mask);
927 return;
929 chip->cores_mask &= pcc->cores_mask;
931 /* now that we have a sane layout, let check the number of cores */
932 cores_max = ctpop64(chip->cores_mask);
933 if (chip->nr_cores > cores_max) {
934 error_setg(errp, "warning: too many cores for chip ! Limit is %d",
935 cores_max);
936 return;
940 static void pnv_chip_instance_init(Object *obj)
942 PNV_CHIP(obj)->xscom_base = PNV_CHIP_GET_CLASS(obj)->xscom_base;
945 static void pnv_chip_core_realize(PnvChip *chip, Error **errp)
947 Error *error = NULL;
948 PnvChipClass *pcc = PNV_CHIP_GET_CLASS(chip);
949 const char *typename = pnv_chip_core_typename(chip);
950 size_t typesize = object_type_get_instance_size(typename);
951 int i, core_hwid;
953 if (!object_class_by_name(typename)) {
954 error_setg(errp, "Unable to find PowerNV CPU Core '%s'", typename);
955 return;
958 /* Cores */
959 pnv_chip_core_sanitize(chip, &error);
960 if (error) {
961 error_propagate(errp, error);
962 return;
965 chip->cores = g_malloc0(typesize * chip->nr_cores);
967 for (i = 0, core_hwid = 0; (core_hwid < sizeof(chip->cores_mask) * 8)
968 && (i < chip->nr_cores); core_hwid++) {
969 char core_name[32];
970 void *pnv_core = chip->cores + i * typesize;
971 uint64_t xscom_core_base;
973 if (!(chip->cores_mask & (1ull << core_hwid))) {
974 continue;
977 object_initialize(pnv_core, typesize, typename);
978 snprintf(core_name, sizeof(core_name), "core[%d]", core_hwid);
979 object_property_add_child(OBJECT(chip), core_name, OBJECT(pnv_core),
980 &error_fatal);
981 object_property_set_int(OBJECT(pnv_core), smp_threads, "nr-threads",
982 &error_fatal);
983 object_property_set_int(OBJECT(pnv_core), core_hwid,
984 CPU_CORE_PROP_CORE_ID, &error_fatal);
985 object_property_set_int(OBJECT(pnv_core),
986 pcc->core_pir(chip, core_hwid),
987 "pir", &error_fatal);
988 object_property_add_const_link(OBJECT(pnv_core), "chip",
989 OBJECT(chip), &error_fatal);
990 object_property_set_bool(OBJECT(pnv_core), true, "realized",
991 &error_fatal);
992 object_unref(OBJECT(pnv_core));
994 /* Each core has an XSCOM MMIO region */
995 if (!pnv_chip_is_power9(chip)) {
996 xscom_core_base = PNV_XSCOM_EX_BASE(core_hwid);
997 } else {
998 xscom_core_base = PNV_XSCOM_P9_EC_BASE(core_hwid);
1001 pnv_xscom_add_subregion(chip, xscom_core_base,
1002 &PNV_CORE(pnv_core)->xscom_regs);
1003 i++;
1007 static void pnv_chip_realize(DeviceState *dev, Error **errp)
1009 PnvChip *chip = PNV_CHIP(dev);
1010 Error *error = NULL;
1012 /* XSCOM bridge */
1013 pnv_xscom_realize(chip, &error);
1014 if (error) {
1015 error_propagate(errp, error);
1016 return;
1018 sysbus_mmio_map(SYS_BUS_DEVICE(chip), 0, PNV_XSCOM_BASE(chip));
1020 /* Cores */
1021 pnv_chip_core_realize(chip, &error);
1022 if (error) {
1023 error_propagate(errp, error);
1024 return;
1028 static Property pnv_chip_properties[] = {
1029 DEFINE_PROP_UINT32("chip-id", PnvChip, chip_id, 0),
1030 DEFINE_PROP_UINT64("ram-start", PnvChip, ram_start, 0),
1031 DEFINE_PROP_UINT64("ram-size", PnvChip, ram_size, 0),
1032 DEFINE_PROP_UINT32("nr-cores", PnvChip, nr_cores, 1),
1033 DEFINE_PROP_UINT64("cores-mask", PnvChip, cores_mask, 0x0),
1034 DEFINE_PROP_END_OF_LIST(),
1037 static void pnv_chip_class_init(ObjectClass *klass, void *data)
1039 DeviceClass *dc = DEVICE_CLASS(klass);
1041 set_bit(DEVICE_CATEGORY_CPU, dc->categories);
1042 dc->realize = pnv_chip_realize;
1043 dc->props = pnv_chip_properties;
1044 dc->desc = "PowerNV Chip";
1047 static ICSState *pnv_ics_get(XICSFabric *xi, int irq)
1049 PnvMachineState *pnv = PNV_MACHINE(xi);
1050 int i;
1052 for (i = 0; i < pnv->num_chips; i++) {
1053 Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
1055 if (ics_valid_irq(&chip8->psi.ics, irq)) {
1056 return &chip8->psi.ics;
1059 return NULL;
1062 static void pnv_ics_resend(XICSFabric *xi)
1064 PnvMachineState *pnv = PNV_MACHINE(xi);
1065 int i;
1067 for (i = 0; i < pnv->num_chips; i++) {
1068 Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
1069 ics_resend(&chip8->psi.ics);
1073 static PowerPCCPU *ppc_get_vcpu_by_pir(int pir)
1075 CPUState *cs;
1077 CPU_FOREACH(cs) {
1078 PowerPCCPU *cpu = POWERPC_CPU(cs);
1079 CPUPPCState *env = &cpu->env;
1081 if (env->spr_cb[SPR_PIR].default_value == pir) {
1082 return cpu;
1086 return NULL;
1089 static ICPState *pnv_icp_get(XICSFabric *xi, int pir)
1091 PowerPCCPU *cpu = ppc_get_vcpu_by_pir(pir);
1093 return cpu ? ICP(cpu->intc) : NULL;
1096 static void pnv_pic_print_info(InterruptStatsProvider *obj,
1097 Monitor *mon)
1099 PnvMachineState *pnv = PNV_MACHINE(obj);
1100 int i;
1101 CPUState *cs;
1103 CPU_FOREACH(cs) {
1104 PowerPCCPU *cpu = POWERPC_CPU(cs);
1106 icp_pic_print_info(ICP(cpu->intc), mon);
1109 for (i = 0; i < pnv->num_chips; i++) {
1110 Pnv8Chip *chip8 = PNV8_CHIP(pnv->chips[i]);
1111 ics_pic_print_info(&chip8->psi.ics, mon);
1115 static void pnv_get_num_chips(Object *obj, Visitor *v, const char *name,
1116 void *opaque, Error **errp)
1118 visit_type_uint32(v, name, &PNV_MACHINE(obj)->num_chips, errp);
1121 static void pnv_set_num_chips(Object *obj, Visitor *v, const char *name,
1122 void *opaque, Error **errp)
1124 PnvMachineState *pnv = PNV_MACHINE(obj);
1125 uint32_t num_chips;
1126 Error *local_err = NULL;
1128 visit_type_uint32(v, name, &num_chips, &local_err);
1129 if (local_err) {
1130 error_propagate(errp, local_err);
1131 return;
1135 * TODO: should we decide on how many chips we can create based
1136 * on #cores and Venice vs. Murano vs. Naples chip type etc...,
1138 if (!is_power_of_2(num_chips) || num_chips > 4) {
1139 error_setg(errp, "invalid number of chips: '%d'", num_chips);
1140 return;
1143 pnv->num_chips = num_chips;
1146 static void pnv_machine_instance_init(Object *obj)
1148 PnvMachineState *pnv = PNV_MACHINE(obj);
1149 pnv->num_chips = 1;
1152 static void pnv_machine_class_props_init(ObjectClass *oc)
1154 object_class_property_add(oc, "num-chips", "uint32",
1155 pnv_get_num_chips, pnv_set_num_chips,
1156 NULL, NULL, NULL);
1157 object_class_property_set_description(oc, "num-chips",
1158 "Specifies the number of processor chips",
1159 NULL);
1162 static void pnv_machine_class_init(ObjectClass *oc, void *data)
1164 MachineClass *mc = MACHINE_CLASS(oc);
1165 XICSFabricClass *xic = XICS_FABRIC_CLASS(oc);
1166 InterruptStatsProviderClass *ispc = INTERRUPT_STATS_PROVIDER_CLASS(oc);
1168 mc->desc = "IBM PowerNV (Non-Virtualized)";
1169 mc->init = pnv_init;
1170 mc->reset = pnv_reset;
1171 mc->max_cpus = MAX_CPUS;
1172 mc->default_cpu_type = POWERPC_CPU_TYPE_NAME("power8_v2.0");
1173 mc->block_default_type = IF_IDE; /* Pnv provides a AHCI device for
1174 * storage */
1175 mc->no_parallel = 1;
1176 mc->default_boot_order = NULL;
1177 mc->default_ram_size = 1 * GiB;
1178 xic->icp_get = pnv_icp_get;
1179 xic->ics_get = pnv_ics_get;
1180 xic->ics_resend = pnv_ics_resend;
1181 ispc->print_info = pnv_pic_print_info;
1183 pnv_machine_class_props_init(oc);
1186 #define DEFINE_PNV8_CHIP_TYPE(type, class_initfn) \
1188 .name = type, \
1189 .class_init = class_initfn, \
1190 .parent = TYPE_PNV8_CHIP, \
1193 #define DEFINE_PNV9_CHIP_TYPE(type, class_initfn) \
1195 .name = type, \
1196 .class_init = class_initfn, \
1197 .parent = TYPE_PNV9_CHIP, \
1200 static const TypeInfo types[] = {
1202 .name = TYPE_PNV_MACHINE,
1203 .parent = TYPE_MACHINE,
1204 .instance_size = sizeof(PnvMachineState),
1205 .instance_init = pnv_machine_instance_init,
1206 .class_init = pnv_machine_class_init,
1207 .interfaces = (InterfaceInfo[]) {
1208 { TYPE_XICS_FABRIC },
1209 { TYPE_INTERRUPT_STATS_PROVIDER },
1210 { },
1214 .name = TYPE_PNV_CHIP,
1215 .parent = TYPE_SYS_BUS_DEVICE,
1216 .class_init = pnv_chip_class_init,
1217 .instance_init = pnv_chip_instance_init,
1218 .instance_size = sizeof(PnvChip),
1219 .class_size = sizeof(PnvChipClass),
1220 .abstract = true,
1224 * P9 chip and variants
1227 .name = TYPE_PNV9_CHIP,
1228 .parent = TYPE_PNV_CHIP,
1229 .instance_init = pnv_chip_power9_instance_init,
1230 .instance_size = sizeof(Pnv9Chip),
1232 DEFINE_PNV9_CHIP_TYPE(TYPE_PNV_CHIP_POWER9, pnv_chip_power9_class_init),
1235 * P8 chip and variants
1238 .name = TYPE_PNV8_CHIP,
1239 .parent = TYPE_PNV_CHIP,
1240 .instance_init = pnv_chip_power8_instance_init,
1241 .instance_size = sizeof(Pnv8Chip),
1243 DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8, pnv_chip_power8_class_init),
1244 DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8E, pnv_chip_power8e_class_init),
1245 DEFINE_PNV8_CHIP_TYPE(TYPE_PNV_CHIP_POWER8NVL,
1246 pnv_chip_power8nvl_class_init),
1249 DEFINE_TYPES(types)