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 #include "hw/msi.h"
  45
  46 #include "kvm.h"
  47 #include "kvm_ppc.h"
  48 #include "pci.h"
  49 #include "vga-pci.h"
  50
  51 #include "exec-memory.h"
  52 #include "hw/usb.h"
  53
  54 #include <libfdt.h>
  55
  56 /* SLOF memory layout:
  57  *
  58  * SLOF raw image loaded at 0, copies its romfs right below the flat
  59  * device-tree, then position SLOF itself 31M below that
  60  *
  61  * So we set FW_OVERHEAD to 40MB which should account for all of that
  62  * and more
  63  *
  64  * We load our kernel at 4M, leaving space for SLOF initial image
  65  */
  66 #define FDT_MAX_SIZE            0x10000
  67 #define RTAS_MAX_SIZE           0x10000
  68 #define FW_MAX_SIZE             0x400000
  69 #define FW_FILE_NAME            "slof.bin"
  70 #define FW_OVERHEAD             0x2800000
  71 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
  72
  73 #define MIN_RMA_SLOF            128UL
  74
  75 #define TIMEBASE_FREQ           512000000ULL
  76
  77 #define MAX_CPUS                256
  78 #define XICS_IRQS               1024
  79
  80 #define SPAPR_PCI_BUID          0x800000020000001ULL
  81 #define SPAPR_PCI_MEM_WIN_ADDR  (0x10000000000ULL + 0xA0000000)
  82 #define SPAPR_PCI_MEM_WIN_SIZE  0x20000000
  83 #define SPAPR_PCI_IO_WIN_ADDR   (0x10000000000ULL + 0x80000000)
  84 #define SPAPR_PCI_MSI_WIN_ADDR  (0x10000000000ULL + 0x90000000)
  85
  86 #define PHANDLE_XICP            0x00001111
  87
  88 sPAPREnvironment *spapr;
  89
  90 int spapr_allocate_irq(int hint, enum xics_irq_type type)
  91 {
  92     int irq;
  93
  94     if (hint) {
  95         irq = hint;
  96         /* FIXME: we should probably check for collisions somehow */
  97     } else {
  98         irq = spapr->next_irq++;
  99     }
 100
 101     /* Configure irq type */
 102     if (!xics_get_qirq(spapr->icp, irq)) {
 103         return 0;
 104     }
 105
 106     xics_set_irq_type(spapr->icp, irq, type);
 107
 108     return irq;
 109 }
 110
 111 /* Allocate block of consequtive IRQs, returns a number of the first */
 112 int spapr_allocate_irq_block(int num, enum xics_irq_type type)
 113 {
 114     int first = -1;
 115     int i;
 116
 117     for (i = 0; i < num; ++i) {
 118         int irq;
 119
 120         irq = spapr_allocate_irq(0, type);
 121         if (!irq) {
 122             return -1;
 123         }
 124
 125         if (0 == i) {
 126             first = irq;
 127         }
 128
 129         /* If the above doesn't create a consecutive block then that's
 130          * an internal bug */
 131         assert(irq == (first + i));
 132     }
 133
 134     return first;
 135 }
 136
 137 static int spapr_set_associativity(void *fdt, sPAPREnvironment *spapr)
 138 {
 139     int ret = 0, offset;
 140     CPUPPCState *env;
 141     char cpu_model[32];
 142     int smt = kvmppc_smt_threads();
 143
 144     assert(spapr->cpu_model);
 145
 146     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 147         uint32_t associativity[] = {cpu_to_be32(0x5),
 148                                     cpu_to_be32(0x0),
 149                                     cpu_to_be32(0x0),
 150                                     cpu_to_be32(0x0),
 151                                     cpu_to_be32(env->numa_node),
 152                                     cpu_to_be32(env->cpu_index)};
 153
 154         if ((env->cpu_index % smt) != 0) {
 155             continue;
 156         }
 157
 158         snprintf(cpu_model, 32, "/cpus/%s@%x", spapr->cpu_model,
 159                  env->cpu_index);
 160
 161         offset = fdt_path_offset(fdt, cpu_model);
 162         if (offset < 0) {
 163             return offset;
 164         }
 165
 166         ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
 167                           sizeof(associativity));
 168         if (ret < 0) {
 169             return ret;
 170         }
 171     }
 172     return ret;
 173 }
 174
 175
 176 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
 177                                      size_t maxsize)
 178 {
 179     size_t maxcells = maxsize / sizeof(uint32_t);
 180     int i, j, count;
 181     uint32_t *p = prop;
 182
 183     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
 184         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
 185
 186         if (!sps->page_shift) {
 187             break;
 188         }
 189         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
 190             if (sps->enc[count].page_shift == 0) {
 191                 break;
 192             }
 193         }
 194         if ((p - prop) >= (maxcells - 3 - count * 2)) {
 195             break;
 196         }
 197         *(p++) = cpu_to_be32(sps->page_shift);
 198         *(p++) = cpu_to_be32(sps->slb_enc);
 199         *(p++) = cpu_to_be32(count);
 200         for (j = 0; j < count; j++) {
 201             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
 202             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
 203         }
 204     }
 205
 206     return (p - prop) * sizeof(uint32_t);
 207 }
 208
 209 static void *spapr_create_fdt_skel(const char *cpu_model,
 210                                    target_phys_addr_t rma_size,
 211                                    target_phys_addr_t initrd_base,
 212                                    target_phys_addr_t initrd_size,
 213                                    target_phys_addr_t kernel_size,
 214                                    const char *boot_device,
 215                                    const char *kernel_cmdline,
 216                                    long hash_shift)
 217 {
 218     void *fdt;
 219     CPUPPCState *env;
 220     uint64_t mem_reg_property[2];
 221     uint32_t start_prop = cpu_to_be32(initrd_base);
 222     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
 223     uint32_t pft_size_prop[] = {0, cpu_to_be32(hash_shift)};
 224     char hypertas_prop[] = "hcall-pft\0hcall-term\0hcall-dabr\0hcall-interrupt"
 225         "\0hcall-tce\0hcall-vio\0hcall-splpar\0hcall-bulk";
 226     char qemu_hypertas_prop[] = "hcall-memop1";
 227     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
 228     int i;
 229     char *modelname;
 230     int smt = kvmppc_smt_threads();
 231     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
 232     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
 233     uint32_t associativity[] = {cpu_to_be32(0x4), cpu_to_be32(0x0),
 234                                 cpu_to_be32(0x0), cpu_to_be32(0x0),
 235                                 cpu_to_be32(0x0)};
 236     char mem_name[32];
 237     target_phys_addr_t node0_size, mem_start;
 238
 239 #define _FDT(exp) \
 240     do { \
 241         int ret = (exp);                                           \
 242         if (ret < 0) {                                             \
 243             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
 244                     #exp, fdt_strerror(ret));                      \
 245             exit(1);                                               \
 246         }                                                          \
 247     } while (0)
 248
 249     fdt = g_malloc0(FDT_MAX_SIZE);
 250     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
 251
 252     if (kernel_size) {
 253         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
 254     }
 255     if (initrd_size) {
 256         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
 257     }
 258     _FDT((fdt_finish_reservemap(fdt)));
 259
 260     /* Root node */
 261     _FDT((fdt_begin_node(fdt, "")));
 262     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
 263     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
 264
 265     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
 266     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
 267
 268     /* /chosen */
 269     _FDT((fdt_begin_node(fdt, "chosen")));
 270
 271     /* Set Form1_affinity */
 272     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
 273
 274     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
 275     _FDT((fdt_property(fdt, "linux,initrd-start",
 276                        &start_prop, sizeof(start_prop))));
 277     _FDT((fdt_property(fdt, "linux,initrd-end",
 278                        &end_prop, sizeof(end_prop))));
 279     if (kernel_size) {
 280         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
 281                               cpu_to_be64(kernel_size) };
 282
 283         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
 284     }
 285     _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
 286     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
 287     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
 288     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
 289
 290     _FDT((fdt_end_node(fdt)));
 291
 292     /* memory node(s) */
 293     node0_size = (nb_numa_nodes > 1) ? node_mem[0] : ram_size;
 294     if (rma_size > node0_size) {
 295         rma_size = node0_size;
 296     }
 297
 298     /* RMA */
 299     mem_reg_property[0] = 0;
 300     mem_reg_property[1] = cpu_to_be64(rma_size);
 301     _FDT((fdt_begin_node(fdt, "memory@0")));
 302     _FDT((fdt_property_string(fdt, "device_type", "memory")));
 303     _FDT((fdt_property(fdt, "reg", mem_reg_property,
 304         sizeof(mem_reg_property))));
 305     _FDT((fdt_property(fdt, "ibm,associativity", associativity,
 306         sizeof(associativity))));
 307     _FDT((fdt_end_node(fdt)));
 308
 309     /* RAM: Node 0 */
 310     if (node0_size > rma_size) {
 311         mem_reg_property[0] = cpu_to_be64(rma_size);
 312         mem_reg_property[1] = cpu_to_be64(node0_size - rma_size);
 313
 314         sprintf(mem_name, "memory@" TARGET_FMT_lx, rma_size);
 315         _FDT((fdt_begin_node(fdt, mem_name)));
 316         _FDT((fdt_property_string(fdt, "device_type", "memory")));
 317         _FDT((fdt_property(fdt, "reg", mem_reg_property,
 318                            sizeof(mem_reg_property))));
 319         _FDT((fdt_property(fdt, "ibm,associativity", associativity,
 320                            sizeof(associativity))));
 321         _FDT((fdt_end_node(fdt)));
 322     }
 323
 324     /* RAM: Node 1 and beyond */
 325     mem_start = node0_size;
 326     for (i = 1; i < nb_numa_nodes; i++) {
 327         mem_reg_property[0] = cpu_to_be64(mem_start);
 328         mem_reg_property[1] = cpu_to_be64(node_mem[i]);
 329         associativity[3] = associativity[4] = cpu_to_be32(i);
 330         sprintf(mem_name, "memory@" TARGET_FMT_lx, mem_start);
 331         _FDT((fdt_begin_node(fdt, mem_name)));
 332         _FDT((fdt_property_string(fdt, "device_type", "memory")));
 333         _FDT((fdt_property(fdt, "reg", mem_reg_property,
 334             sizeof(mem_reg_property))));
 335         _FDT((fdt_property(fdt, "ibm,associativity", associativity,
 336             sizeof(associativity))));
 337         _FDT((fdt_end_node(fdt)));
 338         mem_start += node_mem[i];
 339     }
 340
 341     /* cpus */
 342     _FDT((fdt_begin_node(fdt, "cpus")));
 343
 344     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
 345     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
 346
 347     modelname = g_strdup(cpu_model);
 348
 349     for (i = 0; i < strlen(modelname); i++) {
 350         modelname[i] = toupper(modelname[i]);
 351     }
 352
 353     /* This is needed during FDT finalization */
 354     spapr->cpu_model = g_strdup(modelname);
 355
 356     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 357         int index = env->cpu_index;
 358         uint32_t servers_prop[smp_threads];
 359         uint32_t gservers_prop[smp_threads * 2];
 360         char *nodename;
 361         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
 362                            0xffffffff, 0xffffffff};
 363         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
 364         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
 365         uint32_t page_sizes_prop[64];
 366         size_t page_sizes_prop_size;
 367
 368         if ((index % smt) != 0) {
 369             continue;
 370         }
 371
 372         if (asprintf(&nodename, "%s@%x", modelname, index) < 0) {
 373             fprintf(stderr, "Allocation failure\n");
 374             exit(1);
 375         }
 376
 377         _FDT((fdt_begin_node(fdt, nodename)));
 378
 379         free(nodename);
 380
 381         _FDT((fdt_property_cell(fdt, "reg", index)));
 382         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
 383
 384         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
 385         _FDT((fdt_property_cell(fdt, "dcache-block-size",
 386                                 env->dcache_line_size)));
 387         _FDT((fdt_property_cell(fdt, "icache-block-size",
 388                                 env->icache_line_size)));
 389         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
 390         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
 391         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
 392         _FDT((fdt_property(fdt, "ibm,pft-size",
 393                            pft_size_prop, sizeof(pft_size_prop))));
 394         _FDT((fdt_property_string(fdt, "status", "okay")));
 395         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
 396
 397         /* Build interrupt servers and gservers properties */
 398         for (i = 0; i < smp_threads; i++) {
 399             servers_prop[i] = cpu_to_be32(index + i);
 400             /* Hack, direct the group queues back to cpu 0 */
 401             gservers_prop[i*2] = cpu_to_be32(index + i);
 402             gservers_prop[i*2 + 1] = 0;
 403         }
 404         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-server#s",
 405                            servers_prop, sizeof(servers_prop))));
 406         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-gserver#s",
 407                            gservers_prop, sizeof(gservers_prop))));
 408
 409         if (env->mmu_model & POWERPC_MMU_1TSEG) {
 410             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
 411                                segs, sizeof(segs))));
 412         }
 413
 414         /* Advertise VMX/VSX (vector extensions) if available
 415          *   0 / no property == no vector extensions
 416          *   1               == VMX / Altivec available
 417          *   2               == VSX available */
 418         if (env->insns_flags & PPC_ALTIVEC) {
 419             uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
 420
 421             _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
 422         }
 423
 424         /* Advertise DFP (Decimal Floating Point) if available
 425          *   0 / no property == no DFP
 426          *   1               == DFP available */
 427         if (env->insns_flags2 & PPC2_DFP) {
 428             _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
 429         }
 430
 431         page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
 432                                                       sizeof(page_sizes_prop));
 433         if (page_sizes_prop_size) {
 434             _FDT((fdt_property(fdt, "ibm,segment-page-sizes",
 435                                page_sizes_prop, page_sizes_prop_size)));
 436         }
 437
 438         _FDT((fdt_end_node(fdt)));
 439     }
 440
 441     g_free(modelname);
 442
 443     _FDT((fdt_end_node(fdt)));
 444
 445     /* RTAS */
 446     _FDT((fdt_begin_node(fdt, "rtas")));
 447
 448     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas_prop,
 449                        sizeof(hypertas_prop))));
 450     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas_prop,
 451                        sizeof(qemu_hypertas_prop))));
 452
 453     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
 454         refpoints, sizeof(refpoints))));
 455
 456     _FDT((fdt_end_node(fdt)));
 457
 458     /* interrupt controller */
 459     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
 460
 461     _FDT((fdt_property_string(fdt, "device_type",
 462                               "PowerPC-External-Interrupt-Presentation")));
 463     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
 464     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
 465     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
 466                        interrupt_server_ranges_prop,
 467                        sizeof(interrupt_server_ranges_prop))));
 468     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
 469     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
 470     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
 471
 472     _FDT((fdt_end_node(fdt)));
 473
 474     /* vdevice */
 475     _FDT((fdt_begin_node(fdt, "vdevice")));
 476
 477     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
 478     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
 479     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
 480     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
 481     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
 482     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
 483
 484     _FDT((fdt_end_node(fdt)));
 485
 486     _FDT((fdt_end_node(fdt))); /* close root node */
 487     _FDT((fdt_finish(fdt)));
 488
 489     return fdt;
 490 }
 491
 492 static void spapr_finalize_fdt(sPAPREnvironment *spapr,
 493                                target_phys_addr_t fdt_addr,
 494                                target_phys_addr_t rtas_addr,
 495                                target_phys_addr_t rtas_size)
 496 {
 497     int ret;
 498     void *fdt;
 499     sPAPRPHBState *phb;
 500
 501     fdt = g_malloc(FDT_MAX_SIZE);
 502
 503     /* open out the base tree into a temp buffer for the final tweaks */
 504     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
 505
 506     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
 507     if (ret < 0) {
 508         fprintf(stderr, "couldn't setup vio devices in fdt\n");
 509         exit(1);
 510     }
 511
 512     QLIST_FOREACH(phb, &spapr->phbs, list) {
 513         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
 514     }
 515
 516     if (ret < 0) {
 517         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
 518         exit(1);
 519     }
 520
 521     /* RTAS */
 522     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
 523     if (ret < 0) {
 524         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
 525     }
 526
 527     /* Advertise NUMA via ibm,associativity */
 528     if (nb_numa_nodes > 1) {
 529         ret = spapr_set_associativity(fdt, spapr);
 530         if (ret < 0) {
 531             fprintf(stderr, "Couldn't set up NUMA device tree properties\n");
 532         }
 533     }
 534
 535     if (!spapr->has_graphics) {
 536         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
 537     }
 538
 539     _FDT((fdt_pack(fdt)));
 540
 541     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
 542         hw_error("FDT too big ! 0x%x bytes (max is 0x%x)\n",
 543                  fdt_totalsize(fdt), FDT_MAX_SIZE);
 544         exit(1);
 545     }
 546
 547     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
 548
 549     g_free(fdt);
 550 }
 551
 552 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
 553 {
 554     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
 555 }
 556
 557 static void emulate_spapr_hypercall(CPUPPCState *env)
 558 {
 559     env->gpr[3] = spapr_hypercall(env, env->gpr[3], &env->gpr[4]);
 560 }
 561
 562 static void spapr_reset(void *opaque)
 563 {
 564     sPAPREnvironment *spapr = (sPAPREnvironment *)opaque;
 565
 566     /* flush out the hash table */
 567     memset(spapr->htab, 0, spapr->htab_size);
 568
 569     /* Load the fdt */
 570     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
 571                        spapr->rtas_size);
 572
 573     /* Set up the entry state */
 574     first_cpu->gpr[3] = spapr->fdt_addr;
 575     first_cpu->gpr[5] = 0;
 576     first_cpu->halted = 0;
 577     first_cpu->nip = spapr->entry_point;
 578
 579 }
 580
 581 static void spapr_cpu_reset(void *opaque)
 582 {
 583     PowerPCCPU *cpu = opaque;
 584
 585     cpu_reset(CPU(cpu));
 586 }
 587
 588 /* Returns whether we want to use VGA or not */
 589 static int spapr_vga_init(PCIBus *pci_bus)
 590 {
 591     switch (vga_interface_type) {
 592     case VGA_STD:
 593         pci_vga_init(pci_bus);
 594         return 1;
 595     case VGA_NONE:
 596         return 0;
 597     default:
 598         fprintf(stderr, "This vga model is not supported,"
 599                 "currently it only supports -vga std\n");
 600         exit(0);
 601         break;
 602     }
 603 }
 604
 605 /* pSeries LPAR / sPAPR hardware init */
 606 static void ppc_spapr_init(ram_addr_t ram_size,
 607                            const char *boot_device,
 608                            const char *kernel_filename,
 609                            const char *kernel_cmdline,
 610                            const char *initrd_filename,
 611                            const char *cpu_model)
 612 {
 613     PowerPCCPU *cpu;
 614     CPUPPCState *env;
 615     PCIHostState *phb;
 616     int i;
 617     MemoryRegion *sysmem = get_system_memory();
 618     MemoryRegion *ram = g_new(MemoryRegion, 1);
 619     target_phys_addr_t rma_alloc_size, rma_size;
 620     uint32_t initrd_base = 0;
 621     long kernel_size = 0, initrd_size = 0;
 622     long load_limit, rtas_limit, fw_size;
 623     long pteg_shift = 17;
 624     char *filename;
 625
 626     msi_supported = true;
 627
 628     spapr = g_malloc0(sizeof(*spapr));
 629     QLIST_INIT(&spapr->phbs);
 630
 631     cpu_ppc_hypercall = emulate_spapr_hypercall;
 632
 633     /* Allocate RMA if necessary */
 634     rma_alloc_size = kvmppc_alloc_rma("ppc_spapr.rma", sysmem);
 635
 636     if (rma_alloc_size == -1) {
 637         hw_error("qemu: Unable to create RMA\n");
 638         exit(1);
 639     }
 640     if (rma_alloc_size && (rma_alloc_size < ram_size)) {
 641         rma_size = rma_alloc_size;
 642     } else {
 643         rma_size = ram_size;
 644     }
 645
 646     /* We place the device tree and RTAS just below either the top of the RMA,
 647      * or just below 2GB, whichever is lowere, so that it can be
 648      * processed with 32-bit real mode code if necessary */
 649     rtas_limit = MIN(rma_size, 0x80000000);
 650     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
 651     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
 652     load_limit = spapr->fdt_addr - FW_OVERHEAD;
 653
 654     /* init CPUs */
 655     if (cpu_model == NULL) {
 656         cpu_model = kvm_enabled() ? "host" : "POWER7";
 657     }
 658     for (i = 0; i < smp_cpus; i++) {
 659         cpu = cpu_ppc_init(cpu_model);
 660         if (cpu == NULL) {
 661             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
 662             exit(1);
 663         }
 664         env = &cpu->env;
 665
 666         /* Set time-base frequency to 512 MHz */
 667         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
 668         qemu_register_reset(spapr_cpu_reset, cpu);
 669
 670         env->hreset_vector = 0x60;
 671         env->hreset_excp_prefix = 0;
 672         env->gpr[3] = env->cpu_index;
 673     }
 674
 675     /* allocate RAM */
 676     spapr->ram_limit = ram_size;
 677     if (spapr->ram_limit > rma_alloc_size) {
 678         ram_addr_t nonrma_base = rma_alloc_size;
 679         ram_addr_t nonrma_size = spapr->ram_limit - rma_alloc_size;
 680
 681         memory_region_init_ram(ram, "ppc_spapr.ram", nonrma_size);
 682         vmstate_register_ram_global(ram);
 683         memory_region_add_subregion(sysmem, nonrma_base, ram);
 684     }
 685
 686     /* allocate hash page table.  For now we always make this 16mb,
 687      * later we should probably make it scale to the size of guest
 688      * RAM */
 689     spapr->htab_size = 1ULL << (pteg_shift + 7);
 690     spapr->htab = qemu_memalign(spapr->htab_size, spapr->htab_size);
 691
 692     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 693         env->external_htab = spapr->htab;
 694         env->htab_base = -1;
 695         env->htab_mask = spapr->htab_size - 1;
 696
 697         /* Tell KVM that we're in PAPR mode */
 698         env->spr[SPR_SDR1] = (unsigned long)spapr->htab |
 699                              ((pteg_shift + 7) - 18);
 700         env->spr[SPR_HIOR] = 0;
 701
 702         if (kvm_enabled()) {
 703             kvmppc_set_papr(env);
 704         }
 705     }
 706
 707     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
 708     spapr->rtas_size = load_image_targphys(filename, spapr->rtas_addr,
 709                                            rtas_limit - spapr->rtas_addr);
 710     if (spapr->rtas_size < 0) {
 711         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
 712         exit(1);
 713     }
 714     if (spapr->rtas_size > RTAS_MAX_SIZE) {
 715         hw_error("RTAS too big ! 0x%lx bytes (max is 0x%x)\n",
 716                  spapr->rtas_size, RTAS_MAX_SIZE);
 717         exit(1);
 718     }
 719     g_free(filename);
 720
 721
 722     /* Set up Interrupt Controller */
 723     spapr->icp = xics_system_init(XICS_IRQS);
 724     spapr->next_irq = 16;
 725
 726     /* Set up IOMMU */
 727     spapr_iommu_init();
 728
 729     /* Set up VIO bus */
 730     spapr->vio_bus = spapr_vio_bus_init();
 731
 732     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
 733         if (serial_hds[i]) {
 734             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
 735         }
 736     }
 737
 738     /* Set up PCI */
 739     spapr_pci_rtas_init();
 740
 741     spapr_create_phb(spapr, "pci", SPAPR_PCI_BUID,
 742                      SPAPR_PCI_MEM_WIN_ADDR,
 743                      SPAPR_PCI_MEM_WIN_SIZE,
 744                      SPAPR_PCI_IO_WIN_ADDR,
 745                      SPAPR_PCI_MSI_WIN_ADDR);
 746     phb = PCI_HOST_BRIDGE(QLIST_FIRST(&spapr->phbs));
 747
 748     for (i = 0; i < nb_nics; i++) {
 749         NICInfo *nd = &nd_table[i];
 750
 751         if (!nd->model) {
 752             nd->model = g_strdup("ibmveth");
 753         }
 754
 755         if (strcmp(nd->model, "ibmveth") == 0) {
 756             spapr_vlan_create(spapr->vio_bus, nd);
 757         } else {
 758             pci_nic_init_nofail(&nd_table[i], nd->model, NULL);
 759         }
 760     }
 761
 762     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
 763         spapr_vscsi_create(spapr->vio_bus);
 764     }
 765
 766     /* Graphics */
 767     if (spapr_vga_init(phb->bus)) {
 768         spapr->has_graphics = true;
 769     }
 770
 771     if (usb_enabled) {
 772         pci_create_simple(phb->bus, -1, "pci-ohci");
 773         if (spapr->has_graphics) {
 774             usbdevice_create("keyboard");
 775             usbdevice_create("mouse");
 776         }
 777     }
 778
 779     if (rma_size < (MIN_RMA_SLOF << 20)) {
 780         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
 781                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
 782         exit(1);
 783     }
 784
 785     if (kernel_filename) {
 786         uint64_t lowaddr = 0;
 787
 788         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
 789                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
 790         if (kernel_size < 0) {
 791             kernel_size = load_image_targphys(kernel_filename,
 792                                               KERNEL_LOAD_ADDR,
 793                                               load_limit - KERNEL_LOAD_ADDR);
 794         }
 795         if (kernel_size < 0) {
 796             fprintf(stderr, "qemu: could not load kernel '%s'\n",
 797                     kernel_filename);
 798             exit(1);
 799         }
 800
 801         /* load initrd */
 802         if (initrd_filename) {
 803             /* Try to locate the initrd in the gap between the kernel
 804              * and the firmware. Add a bit of space just in case
 805              */
 806             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
 807             initrd_size = load_image_targphys(initrd_filename, initrd_base,
 808                                               load_limit - initrd_base);
 809             if (initrd_size < 0) {
 810                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
 811                         initrd_filename);
 812                 exit(1);
 813             }
 814         } else {
 815             initrd_base = 0;
 816             initrd_size = 0;
 817         }
 818     }
 819
 820     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, FW_FILE_NAME);
 821     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
 822     if (fw_size < 0) {
 823         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
 824         exit(1);
 825     }
 826     g_free(filename);
 827
 828     spapr->entry_point = 0x100;
 829
 830     /* SLOF will startup the secondary CPUs using RTAS */
 831     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 832         env->halted = 1;
 833     }
 834
 835     /* Prepare the device tree */
 836     spapr->fdt_skel = spapr_create_fdt_skel(cpu_model, rma_size,
 837                                             initrd_base, initrd_size,
 838                                             kernel_size,
 839                                             boot_device, kernel_cmdline,
 840                                             pteg_shift + 7);
 841     assert(spapr->fdt_skel != NULL);
 842
 843     qemu_register_reset(spapr_reset, spapr);
 844 }
 845
 846 static QEMUMachine spapr_machine = {
 847     .name = "pseries",
 848     .desc = "pSeries Logical Partition (PAPR compliant)",
 849     .init = ppc_spapr_init,
 850     .max_cpus = MAX_CPUS,
 851     .no_parallel = 1,
 852     .use_scsi = 1,
 853 };
 854
 855 static void spapr_machine_init(void)
 856 {
 857     qemu_register_machine(&spapr_machine);
 858 }
 859
 860 machine_init(spapr_machine_init);