hw/spapr.c

   1 /*
   2  * QEMU PowerPC pSeries Logical Partition (aka sPAPR) hardware System Emulator
   3  *
   4  * Copyright (c) 2004-2007 Fabrice Bellard
   5  * Copyright (c) 2007 Jocelyn Mayer
   6  * Copyright (c) 2010 David Gibson, IBM Corporation.
   7  *
   8  * Permission is hereby granted, free of charge, to any person obtaining a copy
   9  * of this software and associated documentation files (the "Software"), to deal
  10  * in the Software without restriction, including without limitation the rights
  11  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
  12  * copies of the Software, and to permit persons to whom the Software is
  13  * furnished to do so, subject to the following conditions:
  14  *
  15  * The above copyright notice and this permission notice shall be included in
  16  * all copies or substantial portions of the Software.
  17  *
  18  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  19  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  20  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
  21  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  22  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  23  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
  24  * THE SOFTWARE.
  25  *
  26  */
  27 #include "sysemu.h"
  28 #include "hw.h"
  29 #include "elf.h"
  30 #include "net.h"
  31 #include "blockdev.h"
  32 #include "cpus.h"
  33 #include "kvm.h"
  34 #include "kvm_ppc.h"
  35
  36 #include "hw/boards.h"
  37 #include "hw/ppc.h"
  38 #include "hw/loader.h"
  39
  40 #include "hw/spapr.h"
  41 #include "hw/spapr_vio.h"
  42 #include "hw/spapr_pci.h"
  43 #include "hw/xics.h"
  44
  45 #include "kvm.h"
  46 #include "kvm_ppc.h"
  47 #include "pci.h"
  48
  49 #include "exec-memory.h"
  50
  51 #include <libfdt.h>
  52
  53 /* SLOF memory layout:
  54  *
  55  * SLOF raw image loaded at 0, copies its romfs right below the flat
  56  * device-tree, then position SLOF itself 31M below that
  57  *
  58  * So we set FW_OVERHEAD to 40MB which should account for all of that
  59  * and more
  60  *
  61  * We load our kernel at 4M, leaving space for SLOF initial image
  62  */
  63 #define FDT_MAX_SIZE            0x10000
  64 #define RTAS_MAX_SIZE           0x10000
  65 #define FW_MAX_SIZE             0x400000
  66 #define FW_FILE_NAME            "slof.bin"
  67 #define FW_OVERHEAD             0x2800000
  68 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
  69
  70 #define MIN_RMA_SLOF            128UL
  71
  72 #define TIMEBASE_FREQ           512000000ULL
  73
  74 #define MAX_CPUS                256
  75 #define XICS_IRQS               1024
  76
  77 #define SPAPR_PCI_BUID          0x800000020000001ULL
  78 #define SPAPR_PCI_MEM_WIN_ADDR  (0x10000000000ULL + 0xA0000000)
  79 #define SPAPR_PCI_MEM_WIN_SIZE  0x20000000
  80 #define SPAPR_PCI_IO_WIN_ADDR   (0x10000000000ULL + 0x80000000)
  81
  82 #define PHANDLE_XICP            0x00001111
  83
  84 sPAPREnvironment *spapr;
  85
  86 qemu_irq spapr_allocate_irq(uint32_t hint, uint32_t *irq_num,
  87                             enum xics_irq_type type)
  88 {
  89     uint32_t irq;
  90     qemu_irq qirq;
  91
  92     if (hint) {
  93         irq = hint;
  94         /* FIXME: we should probably check for collisions somehow */
  95     } else {
  96         irq = spapr->next_irq++;
  97     }
  98
  99     qirq = xics_assign_irq(spapr->icp, irq, type);
 100     if (!qirq) {
 101         return NULL;
 102     }
 103
 104     if (irq_num) {
 105         *irq_num = irq;
 106     }
 107
 108     return qirq;
 109 }
 110
 111 static int spapr_set_associativity(void *fdt, sPAPREnvironment *spapr)
 112 {
 113     int ret = 0, offset;
 114     CPUPPCState *env;
 115     char cpu_model[32];
 116     int smt = kvmppc_smt_threads();
 117
 118     assert(spapr->cpu_model);
 119
 120     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 121         uint32_t associativity[] = {cpu_to_be32(0x5),
 122                                     cpu_to_be32(0x0),
 123                                     cpu_to_be32(0x0),
 124                                     cpu_to_be32(0x0),
 125                                     cpu_to_be32(env->numa_node),
 126                                     cpu_to_be32(env->cpu_index)};
 127
 128         if ((env->cpu_index % smt) != 0) {
 129             continue;
 130         }
 131
 132         snprintf(cpu_model, 32, "/cpus/%s@%x", spapr->cpu_model,
 133                  env->cpu_index);
 134
 135         offset = fdt_path_offset(fdt, cpu_model);
 136         if (offset < 0) {
 137             return offset;
 138         }
 139
 140         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
 141                           sizeof(associativity));
 142         if (ret < 0) {
 143             return ret;
 144         }
 145     }
 146     return ret;
 147 }
 148
 149 static void *spapr_create_fdt_skel(const char *cpu_model,
 150                                    target_phys_addr_t rma_size,
 151                                    target_phys_addr_t initrd_base,
 152                                    target_phys_addr_t initrd_size,
 153                                    target_phys_addr_t kernel_size,
 154                                    const char *boot_device,
 155                                    const char *kernel_cmdline,
 156                                    long hash_shift)
 157 {
 158     void *fdt;
 159     CPUPPCState *env;
 160     uint64_t mem_reg_property[2];
 161     uint32_t start_prop = cpu_to_be32(initrd_base);
 162     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
 163     uint32_t pft_size_prop[] = {0, cpu_to_be32(hash_shift)};
 164     char hypertas_prop[] = "hcall-pft\0hcall-term\0hcall-dabr\0hcall-interrupt"
 165         "\0hcall-tce\0hcall-vio\0hcall-splpar\0hcall-bulk";
 166     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
 167     int i;
 168     char *modelname;
 169     int smt = kvmppc_smt_threads();
 170     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
 171     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
 172     uint32_t associativity[] = {cpu_to_be32(0x4), cpu_to_be32(0x0),
 173                                 cpu_to_be32(0x0), cpu_to_be32(0x0),
 174                                 cpu_to_be32(0x0)};
 175     char mem_name[32];
 176     target_phys_addr_t node0_size, mem_start;
 177
 178 #define _FDT(exp) \
 179     do { \
 180         int ret = (exp);                                           \
 181         if (ret < 0) {                                             \
 182             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
 183                     #exp, fdt_strerror(ret));                      \
 184             exit(1);                                               \
 185         }                                                          \
 186     } while (0)
 187
 188     fdt = g_malloc0(FDT_MAX_SIZE);
 189     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
 190
 191     if (kernel_size) {
 192         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
 193     }
 194     if (initrd_size) {
 195         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
 196     }
 197     _FDT((fdt_finish_reservemap(fdt)));
 198
 199     /* Root node */
 200     _FDT((fdt_begin_node(fdt, "")));
 201     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
 202     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
 203
 204     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
 205     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
 206
 207     /* /chosen */
 208     _FDT((fdt_begin_node(fdt, "chosen")));
 209
 210     /* Set Form1_affinity */
 211     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
 212
 213     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
 214     _FDT((fdt_property(fdt, "linux,initrd-start",
 215                        &start_prop, sizeof(start_prop))));
 216     _FDT((fdt_property(fdt, "linux,initrd-end",
 217                        &end_prop, sizeof(end_prop))));
 218     if (kernel_size) {
 219         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
 220                               cpu_to_be64(kernel_size) };
 221
 222         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
 223     }
 224     _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
 225
 226     _FDT((fdt_end_node(fdt)));
 227
 228     /* memory node(s) */
 229     node0_size = (nb_numa_nodes > 1) ? node_mem[0] : ram_size;
 230     if (rma_size > node0_size) {
 231         rma_size = node0_size;
 232     }
 233
 234     /* RMA */
 235     mem_reg_property[0] = 0;
 236     mem_reg_property[1] = cpu_to_be64(rma_size);
 237     _FDT((fdt_begin_node(fdt, "memory@0")));
 238     _FDT((fdt_property_string(fdt, "device_type", "memory")));
 239     _FDT((fdt_property(fdt, "reg", mem_reg_property,
 240         sizeof(mem_reg_property))));
 241     _FDT((fdt_property(fdt, "ibm,associativity", associativity,
 242         sizeof(associativity))));
 243     _FDT((fdt_end_node(fdt)));
 244
 245     /* RAM: Node 0 */
 246     if (node0_size > rma_size) {
 247         mem_reg_property[0] = cpu_to_be64(rma_size);
 248         mem_reg_property[1] = cpu_to_be64(node0_size - rma_size);
 249
 250         sprintf(mem_name, "memory@" TARGET_FMT_lx, rma_size);
 251         _FDT((fdt_begin_node(fdt, mem_name)));
 252         _FDT((fdt_property_string(fdt, "device_type", "memory")));
 253         _FDT((fdt_property(fdt, "reg", mem_reg_property,
 254                            sizeof(mem_reg_property))));
 255         _FDT((fdt_property(fdt, "ibm,associativity", associativity,
 256                            sizeof(associativity))));
 257         _FDT((fdt_end_node(fdt)));
 258     }
 259
 260     /* RAM: Node 1 and beyond */
 261     mem_start = node0_size;
 262     for (i = 1; i < nb_numa_nodes; i++) {
 263         mem_reg_property[0] = cpu_to_be64(mem_start);
 264         mem_reg_property[1] = cpu_to_be64(node_mem[i]);
 265         associativity[3] = associativity[4] = cpu_to_be32(i);
 266         sprintf(mem_name, "memory@" TARGET_FMT_lx, mem_start);
 267         _FDT((fdt_begin_node(fdt, mem_name)));
 268         _FDT((fdt_property_string(fdt, "device_type", "memory")));
 269         _FDT((fdt_property(fdt, "reg", mem_reg_property,
 270             sizeof(mem_reg_property))));
 271         _FDT((fdt_property(fdt, "ibm,associativity", associativity,
 272             sizeof(associativity))));
 273         _FDT((fdt_end_node(fdt)));
 274         mem_start += node_mem[i];
 275     }
 276
 277     /* cpus */
 278     _FDT((fdt_begin_node(fdt, "cpus")));
 279
 280     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
 281     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
 282
 283     modelname = g_strdup(cpu_model);
 284
 285     for (i = 0; i < strlen(modelname); i++) {
 286         modelname[i] = toupper(modelname[i]);
 287     }
 288
 289     /* This is needed during FDT finalization */
 290     spapr->cpu_model = g_strdup(modelname);
 291
 292     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 293         int index = env->cpu_index;
 294         uint32_t servers_prop[smp_threads];
 295         uint32_t gservers_prop[smp_threads * 2];
 296         char *nodename;
 297         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
 298                            0xffffffff, 0xffffffff};
 299         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
 300         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
 301
 302         if ((index % smt) != 0) {
 303             continue;
 304         }
 305
 306         if (asprintf(&nodename, "%s@%x", modelname, index) < 0) {
 307             fprintf(stderr, "Allocation failure\n");
 308             exit(1);
 309         }
 310
 311         _FDT((fdt_begin_node(fdt, nodename)));
 312
 313         free(nodename);
 314
 315         _FDT((fdt_property_cell(fdt, "reg", index)));
 316         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
 317
 318         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
 319         _FDT((fdt_property_cell(fdt, "dcache-block-size",
 320                                 env->dcache_line_size)));
 321         _FDT((fdt_property_cell(fdt, "icache-block-size",
 322                                 env->icache_line_size)));
 323         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
 324         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
 325         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
 326         _FDT((fdt_property(fdt, "ibm,pft-size",
 327                            pft_size_prop, sizeof(pft_size_prop))));
 328         _FDT((fdt_property_string(fdt, "status", "okay")));
 329         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
 330
 331         /* Build interrupt servers and gservers properties */
 332         for (i = 0; i < smp_threads; i++) {
 333             servers_prop[i] = cpu_to_be32(index + i);
 334             /* Hack, direct the group queues back to cpu 0 */
 335             gservers_prop[i*2] = cpu_to_be32(index + i);
 336             gservers_prop[i*2 + 1] = 0;
 337         }
 338         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-server#s",
 339                            servers_prop, sizeof(servers_prop))));
 340         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-gserver#s",
 341                            gservers_prop, sizeof(gservers_prop))));
 342
 343         if (env->mmu_model & POWERPC_MMU_1TSEG) {
 344             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
 345                                segs, sizeof(segs))));
 346         }
 347
 348         /* Advertise VMX/VSX (vector extensions) if available
 349          *   0 / no property == no vector extensions
 350          *   1               == VMX / Altivec available
 351          *   2               == VSX available */
 352         if (env->insns_flags & PPC_ALTIVEC) {
 353             uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
 354
 355             _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
 356         }
 357
 358         /* Advertise DFP (Decimal Floating Point) if available
 359          *   0 / no property == no DFP
 360          *   1               == DFP available */
 361         if (env->insns_flags2 & PPC2_DFP) {
 362             _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
 363         }
 364
 365         _FDT((fdt_end_node(fdt)));
 366     }
 367
 368     g_free(modelname);
 369
 370     _FDT((fdt_end_node(fdt)));
 371
 372     /* RTAS */
 373     _FDT((fdt_begin_node(fdt, "rtas")));
 374
 375     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas_prop,
 376                        sizeof(hypertas_prop))));
 377
 378     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
 379         refpoints, sizeof(refpoints))));
 380
 381     _FDT((fdt_end_node(fdt)));
 382
 383     /* interrupt controller */
 384     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
 385
 386     _FDT((fdt_property_string(fdt, "device_type",
 387                               "PowerPC-External-Interrupt-Presentation")));
 388     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
 389     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
 390     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
 391                        interrupt_server_ranges_prop,
 392                        sizeof(interrupt_server_ranges_prop))));
 393     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
 394     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
 395     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
 396
 397     _FDT((fdt_end_node(fdt)));
 398
 399     /* vdevice */
 400     _FDT((fdt_begin_node(fdt, "vdevice")));
 401
 402     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
 403     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
 404     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
 405     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
 406     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
 407     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
 408
 409     _FDT((fdt_end_node(fdt)));
 410
 411     _FDT((fdt_end_node(fdt))); /* close root node */
 412     _FDT((fdt_finish(fdt)));
 413
 414     return fdt;
 415 }
 416
 417 static void spapr_finalize_fdt(sPAPREnvironment *spapr,
 418                                target_phys_addr_t fdt_addr,
 419                                target_phys_addr_t rtas_addr,
 420                                target_phys_addr_t rtas_size)
 421 {
 422     int ret;
 423     void *fdt;
 424     sPAPRPHBState *phb;
 425
 426     fdt = g_malloc(FDT_MAX_SIZE);
 427
 428     /* open out the base tree into a temp buffer for the final tweaks */
 429     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
 430
 431     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
 432     if (ret < 0) {
 433         fprintf(stderr, "couldn't setup vio devices in fdt\n");
 434         exit(1);
 435     }
 436
 437     QLIST_FOREACH(phb, &spapr->phbs, list) {
 438         ret = spapr_populate_pci_devices(phb, PHANDLE_XICP, fdt);
 439     }
 440
 441     if (ret < 0) {
 442         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
 443         exit(1);
 444     }
 445
 446     /* RTAS */
 447     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
 448     if (ret < 0) {
 449         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
 450     }
 451
 452     /* Advertise NUMA via ibm,associativity */
 453     if (nb_numa_nodes > 1) {
 454         ret = spapr_set_associativity(fdt, spapr);
 455         if (ret < 0) {
 456             fprintf(stderr, "Couldn't set up NUMA device tree properties\n");
 457         }
 458     }
 459
 460     spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
 461
 462     _FDT((fdt_pack(fdt)));
 463
 464     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
 465         hw_error("FDT too big ! 0x%x bytes (max is 0x%x)\n",
 466                  fdt_totalsize(fdt), FDT_MAX_SIZE);
 467         exit(1);
 468     }
 469
 470     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
 471
 472     g_free(fdt);
 473 }
 474
 475 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
 476 {
 477     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
 478 }
 479
 480 static void emulate_spapr_hypercall(CPUPPCState *env)
 481 {
 482     env->gpr[3] = spapr_hypercall(env, env->gpr[3], &env->gpr[4]);
 483 }
 484
 485 static void spapr_reset(void *opaque)
 486 {
 487     sPAPREnvironment *spapr = (sPAPREnvironment *)opaque;
 488
 489     fprintf(stderr, "sPAPR reset\n");
 490
 491     /* flush out the hash table */
 492     memset(spapr->htab, 0, spapr->htab_size);
 493
 494     /* Load the fdt */
 495     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
 496                        spapr->rtas_size);
 497
 498     /* Set up the entry state */
 499     first_cpu->gpr[3] = spapr->fdt_addr;
 500     first_cpu->gpr[5] = 0;
 501     first_cpu->halted = 0;
 502     first_cpu->nip = spapr->entry_point;
 503
 504 }
 505
 506 static void spapr_cpu_reset(void *opaque)
 507 {
 508     PowerPCCPU *cpu = opaque;
 509
 510     cpu_reset(CPU(cpu));
 511 }
 512
 513 /* pSeries LPAR / sPAPR hardware init */
 514 static void ppc_spapr_init(ram_addr_t ram_size,
 515                            const char *boot_device,
 516                            const char *kernel_filename,
 517                            const char *kernel_cmdline,
 518                            const char *initrd_filename,
 519                            const char *cpu_model)
 520 {
 521     PowerPCCPU *cpu;
 522     CPUPPCState *env;
 523     int i;
 524     MemoryRegion *sysmem = get_system_memory();
 525     MemoryRegion *ram = g_new(MemoryRegion, 1);
 526     target_phys_addr_t rma_alloc_size, rma_size;
 527     uint32_t initrd_base = 0;
 528     long kernel_size = 0, initrd_size = 0;
 529     long load_limit, rtas_limit, fw_size;
 530     long pteg_shift = 17;
 531     char *filename;
 532
 533     spapr = g_malloc0(sizeof(*spapr));
 534     QLIST_INIT(&spapr->phbs);
 535
 536     cpu_ppc_hypercall = emulate_spapr_hypercall;
 537
 538     /* Allocate RMA if necessary */
 539     rma_alloc_size = kvmppc_alloc_rma("ppc_spapr.rma", sysmem);
 540
 541     if (rma_alloc_size == -1) {
 542         hw_error("qemu: Unable to create RMA\n");
 543         exit(1);
 544     }
 545     if (rma_alloc_size && (rma_alloc_size < ram_size)) {
 546         rma_size = rma_alloc_size;
 547     } else {
 548         rma_size = ram_size;
 549     }
 550
 551     /* We place the device tree and RTAS just below either the top of the RMA,
 552      * or just below 2GB, whichever is lowere, so that it can be
 553      * processed with 32-bit real mode code if necessary */
 554     rtas_limit = MIN(rma_size, 0x80000000);
 555     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
 556     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
 557     load_limit = spapr->fdt_addr - FW_OVERHEAD;
 558
 559     /* init CPUs */
 560     if (cpu_model == NULL) {
 561         cpu_model = kvm_enabled() ? "host" : "POWER7";
 562     }
 563     for (i = 0; i < smp_cpus; i++) {
 564         cpu = cpu_ppc_init(cpu_model);
 565         if (cpu == NULL) {
 566             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
 567             exit(1);
 568         }
 569         env = &cpu->env;
 570
 571         /* Set time-base frequency to 512 MHz */
 572         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
 573         qemu_register_reset(spapr_cpu_reset, cpu);
 574
 575         env->hreset_vector = 0x60;
 576         env->hreset_excp_prefix = 0;
 577         env->gpr[3] = env->cpu_index;
 578     }
 579
 580     /* allocate RAM */
 581     spapr->ram_limit = ram_size;
 582     if (spapr->ram_limit > rma_alloc_size) {
 583         ram_addr_t nonrma_base = rma_alloc_size;
 584         ram_addr_t nonrma_size = spapr->ram_limit - rma_alloc_size;
 585
 586         memory_region_init_ram(ram, "ppc_spapr.ram", nonrma_size);
 587         vmstate_register_ram_global(ram);
 588         memory_region_add_subregion(sysmem, nonrma_base, ram);
 589     }
 590
 591     /* allocate hash page table.  For now we always make this 16mb,
 592      * later we should probably make it scale to the size of guest
 593      * RAM */
 594     spapr->htab_size = 1ULL << (pteg_shift + 7);
 595     spapr->htab = qemu_memalign(spapr->htab_size, spapr->htab_size);
 596
 597     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 598         env->external_htab = spapr->htab;
 599         env->htab_base = -1;
 600         env->htab_mask = spapr->htab_size - 1;
 601
 602         /* Tell KVM that we're in PAPR mode */
 603         env->spr[SPR_SDR1] = (unsigned long)spapr->htab |
 604                              ((pteg_shift + 7) - 18);
 605         env->spr[SPR_HIOR] = 0;
 606
 607         if (kvm_enabled()) {
 608             kvmppc_set_papr(env);
 609         }
 610     }
 611
 612     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
 613     spapr->rtas_size = load_image_targphys(filename, spapr->rtas_addr,
 614                                            rtas_limit - spapr->rtas_addr);
 615     if (spapr->rtas_size < 0) {
 616         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
 617         exit(1);
 618     }
 619     if (spapr->rtas_size > RTAS_MAX_SIZE) {
 620         hw_error("RTAS too big ! 0x%lx bytes (max is 0x%x)\n",
 621                  spapr->rtas_size, RTAS_MAX_SIZE);
 622         exit(1);
 623     }
 624     g_free(filename);
 625
 626
 627     /* Set up Interrupt Controller */
 628     spapr->icp = xics_system_init(XICS_IRQS);
 629     spapr->next_irq = 16;
 630
 631     /* Set up VIO bus */
 632     spapr->vio_bus = spapr_vio_bus_init();
 633
 634     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
 635         if (serial_hds[i]) {
 636             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
 637         }
 638     }
 639
 640     /* Set up PCI */
 641     spapr_create_phb(spapr, "pci", SPAPR_PCI_BUID,
 642                      SPAPR_PCI_MEM_WIN_ADDR,
 643                      SPAPR_PCI_MEM_WIN_SIZE,
 644                      SPAPR_PCI_IO_WIN_ADDR);
 645
 646     for (i = 0; i < nb_nics; i++) {
 647         NICInfo *nd = &nd_table[i];
 648
 649         if (!nd->model) {
 650             nd->model = g_strdup("ibmveth");
 651         }
 652
 653         if (strcmp(nd->model, "ibmveth") == 0) {
 654             spapr_vlan_create(spapr->vio_bus, nd);
 655         } else {
 656             pci_nic_init_nofail(&nd_table[i], nd->model, NULL);
 657         }
 658     }
 659
 660     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
 661         spapr_vscsi_create(spapr->vio_bus);
 662     }
 663
 664     if (rma_size < (MIN_RMA_SLOF << 20)) {
 665         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
 666                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
 667         exit(1);
 668     }
 669
 670     fprintf(stderr, "sPAPR memory map:\n");
 671     fprintf(stderr, "RTAS                 : 0x%08lx..%08lx\n",
 672             (unsigned long)spapr->rtas_addr,
 673             (unsigned long)(spapr->rtas_addr + spapr->rtas_size - 1));
 674     fprintf(stderr, "FDT                  : 0x%08lx..%08lx\n",
 675             (unsigned long)spapr->fdt_addr,
 676             (unsigned long)(spapr->fdt_addr + FDT_MAX_SIZE - 1));
 677
 678     if (kernel_filename) {
 679         uint64_t lowaddr = 0;
 680
 681         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
 682                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
 683         if (kernel_size < 0) {
 684             kernel_size = load_image_targphys(kernel_filename,
 685                                               KERNEL_LOAD_ADDR,
 686                                               load_limit - KERNEL_LOAD_ADDR);
 687         }
 688         if (kernel_size < 0) {
 689             fprintf(stderr, "qemu: could not load kernel '%s'\n",
 690                     kernel_filename);
 691             exit(1);
 692         }
 693         fprintf(stderr, "Kernel               : 0x%08x..%08lx\n",
 694                 KERNEL_LOAD_ADDR, KERNEL_LOAD_ADDR + kernel_size - 1);
 695
 696         /* load initrd */
 697         if (initrd_filename) {
 698             /* Try to locate the initrd in the gap between the kernel
 699              * and the firmware. Add a bit of space just in case
 700              */
 701             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
 702             initrd_size = load_image_targphys(initrd_filename, initrd_base,
 703                                               load_limit - initrd_base);
 704             if (initrd_size < 0) {
 705                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
 706                         initrd_filename);
 707                 exit(1);
 708             }
 709             fprintf(stderr, "Ramdisk              : 0x%08lx..%08lx\n",
 710                     (long)initrd_base, (long)(initrd_base + initrd_size - 1));
 711         } else {
 712             initrd_base = 0;
 713             initrd_size = 0;
 714         }
 715     }
 716
 717     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, FW_FILE_NAME);
 718     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
 719     if (fw_size < 0) {
 720         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
 721         exit(1);
 722     }
 723     g_free(filename);
 724     fprintf(stderr, "Firmware load        : 0x%08x..%08lx\n",
 725             0, fw_size);
 726     fprintf(stderr, "Firmware runtime     : 0x%08lx..%08lx\n",
 727             load_limit, (unsigned long)spapr->fdt_addr);
 728
 729     spapr->entry_point = 0x100;
 730
 731     /* SLOF will startup the secondary CPUs using RTAS */
 732     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 733         env->halted = 1;
 734     }
 735
 736     /* Prepare the device tree */
 737     spapr->fdt_skel = spapr_create_fdt_skel(cpu_model, rma_size,
 738                                             initrd_base, initrd_size,
 739                                             kernel_size,
 740                                             boot_device, kernel_cmdline,
 741                                             pteg_shift + 7);
 742     assert(spapr->fdt_skel != NULL);
 743
 744     qemu_register_reset(spapr_reset, spapr);
 745 }
 746
 747 static QEMUMachine spapr_machine = {
 748     .name = "pseries",
 749     .desc = "pSeries Logical Partition (PAPR compliant)",
 750     .init = ppc_spapr_init,
 751     .max_cpus = MAX_CPUS,
 752     .no_parallel = 1,
 753     .use_scsi = 1,
 754 };
 755
 756 static void spapr_machine_init(void)
 757 {
 758     qemu_register_machine(&spapr_machine);
 759 }
 760
 761 machine_init(spapr_machine_init);