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/pci/msi.h"
  45
  46 #include "kvm.h"
  47 #include "kvm_ppc.h"
  48 #include "pci/pci.h"
  49
  50 #include "exec-memory.h"
  51 #include "hw/usb.h"
  52
  53 #include <libfdt.h>
  54
  55 /* SLOF memory layout:
  56  *
  57  * SLOF raw image loaded at 0, copies its romfs right below the flat
  58  * device-tree, then position SLOF itself 31M below that
  59  *
  60  * So we set FW_OVERHEAD to 40MB which should account for all of that
  61  * and more
  62  *
  63  * We load our kernel at 4M, leaving space for SLOF initial image
  64  */
  65 #define FDT_MAX_SIZE            0x10000
  66 #define RTAS_MAX_SIZE           0x10000
  67 #define FW_MAX_SIZE             0x400000
  68 #define FW_FILE_NAME            "slof.bin"
  69 #define FW_OVERHEAD             0x2800000
  70 #define KERNEL_LOAD_ADDR        FW_MAX_SIZE
  71
  72 #define MIN_RMA_SLOF            128UL
  73
  74 #define TIMEBASE_FREQ           512000000ULL
  75
  76 #define MAX_CPUS                256
  77 #define XICS_IRQS               1024
  78
  79 #define SPAPR_PCI_BUID          0x800000020000001ULL
  80 #define SPAPR_PCI_MEM_WIN_ADDR  (0x10000000000ULL + 0xA0000000)
  81 #define SPAPR_PCI_MEM_WIN_SIZE  0x20000000
  82 #define SPAPR_PCI_IO_WIN_ADDR   (0x10000000000ULL + 0x80000000)
  83 #define SPAPR_PCI_MSI_WIN_ADDR  (0x10000000000ULL + 0x90000000)
  84
  85 #define PHANDLE_XICP            0x00001111
  86
  87 #define HTAB_SIZE(spapr)        (1ULL << ((spapr)->htab_shift))
  88
  89 sPAPREnvironment *spapr;
  90
  91 int spapr_allocate_irq(int hint, bool lsi)
  92 {
  93     int irq;
  94
  95     if (hint) {
  96         irq = hint;
  97         /* FIXME: we should probably check for collisions somehow */
  98     } else {
  99         irq = spapr->next_irq++;
 100     }
 101
 102     /* Configure irq type */
 103     if (!xics_get_qirq(spapr->icp, irq)) {
 104         return 0;
 105     }
 106
 107     xics_set_irq_type(spapr->icp, irq, lsi);
 108
 109     return irq;
 110 }
 111
 112 /* Allocate block of consequtive IRQs, returns a number of the first */
 113 int spapr_allocate_irq_block(int num, bool lsi)
 114 {
 115     int first = -1;
 116     int i;
 117
 118     for (i = 0; i < num; ++i) {
 119         int irq;
 120
 121         irq = spapr_allocate_irq(0, lsi);
 122         if (!irq) {
 123             return -1;
 124         }
 125
 126         if (0 == i) {
 127             first = irq;
 128         }
 129
 130         /* If the above doesn't create a consecutive block then that's
 131          * an internal bug */
 132         assert(irq == (first + i));
 133     }
 134
 135     return first;
 136 }
 137
 138 static int spapr_fixup_cpu_dt(void *fdt, sPAPREnvironment *spapr)
 139 {
 140     int ret = 0, offset;
 141     CPUPPCState *env;
 142     char cpu_model[32];
 143     int smt = kvmppc_smt_threads();
 144     uint32_t pft_size_prop[] = {0, cpu_to_be32(spapr->htab_shift)};
 145
 146     assert(spapr->cpu_model);
 147
 148     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 149         uint32_t associativity[] = {cpu_to_be32(0x5),
 150                                     cpu_to_be32(0x0),
 151                                     cpu_to_be32(0x0),
 152                                     cpu_to_be32(0x0),
 153                                     cpu_to_be32(env->numa_node),
 154                                     cpu_to_be32(env->cpu_index)};
 155
 156         if ((env->cpu_index % smt) != 0) {
 157             continue;
 158         }
 159
 160         snprintf(cpu_model, 32, "/cpus/%s@%x", spapr->cpu_model,
 161                  env->cpu_index);
 162
 163         offset = fdt_path_offset(fdt, cpu_model);
 164         if (offset < 0) {
 165             return offset;
 166         }
 167
 168         if (nb_numa_nodes > 1) {
 169             ret = fdt_setprop(fdt, offset, "ibm,associativity", associativity,
 170                               sizeof(associativity));
 171             if (ret < 0) {
 172                 return ret;
 173             }
 174         }
 175
 176         ret = fdt_setprop(fdt, offset, "ibm,pft-size",
 177                           pft_size_prop, sizeof(pft_size_prop));
 178         if (ret < 0) {
 179             return ret;
 180         }
 181     }
 182     return ret;
 183 }
 184
 185
 186 static size_t create_page_sizes_prop(CPUPPCState *env, uint32_t *prop,
 187                                      size_t maxsize)
 188 {
 189     size_t maxcells = maxsize / sizeof(uint32_t);
 190     int i, j, count;
 191     uint32_t *p = prop;
 192
 193     for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
 194         struct ppc_one_seg_page_size *sps = &env->sps.sps[i];
 195
 196         if (!sps->page_shift) {
 197             break;
 198         }
 199         for (count = 0; count < PPC_PAGE_SIZES_MAX_SZ; count++) {
 200             if (sps->enc[count].page_shift == 0) {
 201                 break;
 202             }
 203         }
 204         if ((p - prop) >= (maxcells - 3 - count * 2)) {
 205             break;
 206         }
 207         *(p++) = cpu_to_be32(sps->page_shift);
 208         *(p++) = cpu_to_be32(sps->slb_enc);
 209         *(p++) = cpu_to_be32(count);
 210         for (j = 0; j < count; j++) {
 211             *(p++) = cpu_to_be32(sps->enc[j].page_shift);
 212             *(p++) = cpu_to_be32(sps->enc[j].pte_enc);
 213         }
 214     }
 215
 216     return (p - prop) * sizeof(uint32_t);
 217 }
 218
 219 #define _FDT(exp) \
 220     do { \
 221         int ret = (exp);                                           \
 222         if (ret < 0) {                                             \
 223             fprintf(stderr, "qemu: error creating device tree: %s: %s\n", \
 224                     #exp, fdt_strerror(ret));                      \
 225             exit(1);                                               \
 226         }                                                          \
 227     } while (0)
 228
 229
 230 static void *spapr_create_fdt_skel(const char *cpu_model,
 231                                    hwaddr initrd_base,
 232                                    hwaddr initrd_size,
 233                                    hwaddr kernel_size,
 234                                    const char *boot_device,
 235                                    const char *kernel_cmdline,
 236                                    uint32_t epow_irq)
 237 {
 238     void *fdt;
 239     CPUPPCState *env;
 240     uint32_t start_prop = cpu_to_be32(initrd_base);
 241     uint32_t end_prop = cpu_to_be32(initrd_base + initrd_size);
 242     char hypertas_prop[] = "hcall-pft\0hcall-term\0hcall-dabr\0hcall-interrupt"
 243         "\0hcall-tce\0hcall-vio\0hcall-splpar\0hcall-bulk";
 244     char qemu_hypertas_prop[] = "hcall-memop1";
 245     uint32_t refpoints[] = {cpu_to_be32(0x4), cpu_to_be32(0x4)};
 246     uint32_t interrupt_server_ranges_prop[] = {0, cpu_to_be32(smp_cpus)};
 247     char *modelname;
 248     int i, smt = kvmppc_smt_threads();
 249     unsigned char vec5[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x80};
 250
 251     fdt = g_malloc0(FDT_MAX_SIZE);
 252     _FDT((fdt_create(fdt, FDT_MAX_SIZE)));
 253
 254     if (kernel_size) {
 255         _FDT((fdt_add_reservemap_entry(fdt, KERNEL_LOAD_ADDR, kernel_size)));
 256     }
 257     if (initrd_size) {
 258         _FDT((fdt_add_reservemap_entry(fdt, initrd_base, initrd_size)));
 259     }
 260     _FDT((fdt_finish_reservemap(fdt)));
 261
 262     /* Root node */
 263     _FDT((fdt_begin_node(fdt, "")));
 264     _FDT((fdt_property_string(fdt, "device_type", "chrp")));
 265     _FDT((fdt_property_string(fdt, "model", "IBM pSeries (emulated by qemu)")));
 266
 267     _FDT((fdt_property_cell(fdt, "#address-cells", 0x2)));
 268     _FDT((fdt_property_cell(fdt, "#size-cells", 0x2)));
 269
 270     /* /chosen */
 271     _FDT((fdt_begin_node(fdt, "chosen")));
 272
 273     /* Set Form1_affinity */
 274     _FDT((fdt_property(fdt, "ibm,architecture-vec-5", vec5, sizeof(vec5))));
 275
 276     _FDT((fdt_property_string(fdt, "bootargs", kernel_cmdline)));
 277     _FDT((fdt_property(fdt, "linux,initrd-start",
 278                        &start_prop, sizeof(start_prop))));
 279     _FDT((fdt_property(fdt, "linux,initrd-end",
 280                        &end_prop, sizeof(end_prop))));
 281     if (kernel_size) {
 282         uint64_t kprop[2] = { cpu_to_be64(KERNEL_LOAD_ADDR),
 283                               cpu_to_be64(kernel_size) };
 284
 285         _FDT((fdt_property(fdt, "qemu,boot-kernel", &kprop, sizeof(kprop))));
 286     }
 287     _FDT((fdt_property_string(fdt, "qemu,boot-device", boot_device)));
 288     _FDT((fdt_property_cell(fdt, "qemu,graphic-width", graphic_width)));
 289     _FDT((fdt_property_cell(fdt, "qemu,graphic-height", graphic_height)));
 290     _FDT((fdt_property_cell(fdt, "qemu,graphic-depth", graphic_depth)));
 291
 292     _FDT((fdt_end_node(fdt)));
 293
 294     /* cpus */
 295     _FDT((fdt_begin_node(fdt, "cpus")));
 296
 297     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
 298     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
 299
 300     modelname = g_strdup(cpu_model);
 301
 302     for (i = 0; i < strlen(modelname); i++) {
 303         modelname[i] = toupper(modelname[i]);
 304     }
 305
 306     /* This is needed during FDT finalization */
 307     spapr->cpu_model = g_strdup(modelname);
 308
 309     for (env = first_cpu; env != NULL; env = env->next_cpu) {
 310         int index = env->cpu_index;
 311         uint32_t servers_prop[smp_threads];
 312         uint32_t gservers_prop[smp_threads * 2];
 313         char *nodename;
 314         uint32_t segs[] = {cpu_to_be32(28), cpu_to_be32(40),
 315                            0xffffffff, 0xffffffff};
 316         uint32_t tbfreq = kvm_enabled() ? kvmppc_get_tbfreq() : TIMEBASE_FREQ;
 317         uint32_t cpufreq = kvm_enabled() ? kvmppc_get_clockfreq() : 1000000000;
 318         uint32_t page_sizes_prop[64];
 319         size_t page_sizes_prop_size;
 320
 321         if ((index % smt) != 0) {
 322             continue;
 323         }
 324
 325         if (asprintf(&nodename, "%s@%x", modelname, index) < 0) {
 326             fprintf(stderr, "Allocation failure\n");
 327             exit(1);
 328         }
 329
 330         _FDT((fdt_begin_node(fdt, nodename)));
 331
 332         free(nodename);
 333
 334         _FDT((fdt_property_cell(fdt, "reg", index)));
 335         _FDT((fdt_property_string(fdt, "device_type", "cpu")));
 336
 337         _FDT((fdt_property_cell(fdt, "cpu-version", env->spr[SPR_PVR])));
 338         _FDT((fdt_property_cell(fdt, "dcache-block-size",
 339                                 env->dcache_line_size)));
 340         _FDT((fdt_property_cell(fdt, "icache-block-size",
 341                                 env->icache_line_size)));
 342         _FDT((fdt_property_cell(fdt, "timebase-frequency", tbfreq)));
 343         _FDT((fdt_property_cell(fdt, "clock-frequency", cpufreq)));
 344         _FDT((fdt_property_cell(fdt, "ibm,slb-size", env->slb_nr)));
 345         _FDT((fdt_property_string(fdt, "status", "okay")));
 346         _FDT((fdt_property(fdt, "64-bit", NULL, 0)));
 347
 348         /* Build interrupt servers and gservers properties */
 349         for (i = 0; i < smp_threads; i++) {
 350             servers_prop[i] = cpu_to_be32(index + i);
 351             /* Hack, direct the group queues back to cpu 0 */
 352             gservers_prop[i*2] = cpu_to_be32(index + i);
 353             gservers_prop[i*2 + 1] = 0;
 354         }
 355         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-server#s",
 356                            servers_prop, sizeof(servers_prop))));
 357         _FDT((fdt_property(fdt, "ibm,ppc-interrupt-gserver#s",
 358                            gservers_prop, sizeof(gservers_prop))));
 359
 360         if (env->mmu_model & POWERPC_MMU_1TSEG) {
 361             _FDT((fdt_property(fdt, "ibm,processor-segment-sizes",
 362                                segs, sizeof(segs))));
 363         }
 364
 365         /* Advertise VMX/VSX (vector extensions) if available
 366          *   0 / no property == no vector extensions
 367          *   1               == VMX / Altivec available
 368          *   2               == VSX available */
 369         if (env->insns_flags & PPC_ALTIVEC) {
 370             uint32_t vmx = (env->insns_flags2 & PPC2_VSX) ? 2 : 1;
 371
 372             _FDT((fdt_property_cell(fdt, "ibm,vmx", vmx)));
 373         }
 374
 375         /* Advertise DFP (Decimal Floating Point) if available
 376          *   0 / no property == no DFP
 377          *   1               == DFP available */
 378         if (env->insns_flags2 & PPC2_DFP) {
 379             _FDT((fdt_property_cell(fdt, "ibm,dfp", 1)));
 380         }
 381
 382         page_sizes_prop_size = create_page_sizes_prop(env, page_sizes_prop,
 383                                                       sizeof(page_sizes_prop));
 384         if (page_sizes_prop_size) {
 385             _FDT((fdt_property(fdt, "ibm,segment-page-sizes",
 386                                page_sizes_prop, page_sizes_prop_size)));
 387         }
 388
 389         _FDT((fdt_end_node(fdt)));
 390     }
 391
 392     g_free(modelname);
 393
 394     _FDT((fdt_end_node(fdt)));
 395
 396     /* RTAS */
 397     _FDT((fdt_begin_node(fdt, "rtas")));
 398
 399     _FDT((fdt_property(fdt, "ibm,hypertas-functions", hypertas_prop,
 400                        sizeof(hypertas_prop))));
 401     _FDT((fdt_property(fdt, "qemu,hypertas-functions", qemu_hypertas_prop,
 402                        sizeof(qemu_hypertas_prop))));
 403
 404     _FDT((fdt_property(fdt, "ibm,associativity-reference-points",
 405         refpoints, sizeof(refpoints))));
 406
 407     _FDT((fdt_property_cell(fdt, "rtas-error-log-max", RTAS_ERROR_LOG_MAX)));
 408
 409     _FDT((fdt_end_node(fdt)));
 410
 411     /* interrupt controller */
 412     _FDT((fdt_begin_node(fdt, "interrupt-controller")));
 413
 414     _FDT((fdt_property_string(fdt, "device_type",
 415                               "PowerPC-External-Interrupt-Presentation")));
 416     _FDT((fdt_property_string(fdt, "compatible", "IBM,ppc-xicp")));
 417     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
 418     _FDT((fdt_property(fdt, "ibm,interrupt-server-ranges",
 419                        interrupt_server_ranges_prop,
 420                        sizeof(interrupt_server_ranges_prop))));
 421     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 2)));
 422     _FDT((fdt_property_cell(fdt, "linux,phandle", PHANDLE_XICP)));
 423     _FDT((fdt_property_cell(fdt, "phandle", PHANDLE_XICP)));
 424
 425     _FDT((fdt_end_node(fdt)));
 426
 427     /* vdevice */
 428     _FDT((fdt_begin_node(fdt, "vdevice")));
 429
 430     _FDT((fdt_property_string(fdt, "device_type", "vdevice")));
 431     _FDT((fdt_property_string(fdt, "compatible", "IBM,vdevice")));
 432     _FDT((fdt_property_cell(fdt, "#address-cells", 0x1)));
 433     _FDT((fdt_property_cell(fdt, "#size-cells", 0x0)));
 434     _FDT((fdt_property_cell(fdt, "#interrupt-cells", 0x2)));
 435     _FDT((fdt_property(fdt, "interrupt-controller", NULL, 0)));
 436
 437     _FDT((fdt_end_node(fdt)));
 438
 439     /* event-sources */
 440     spapr_events_fdt_skel(fdt, epow_irq);
 441
 442     _FDT((fdt_end_node(fdt))); /* close root node */
 443     _FDT((fdt_finish(fdt)));
 444
 445     return fdt;
 446 }
 447
 448 static int spapr_populate_memory(sPAPREnvironment *spapr, void *fdt)
 449 {
 450     uint32_t associativity[] = {cpu_to_be32(0x4), cpu_to_be32(0x0),
 451                                 cpu_to_be32(0x0), cpu_to_be32(0x0),
 452                                 cpu_to_be32(0x0)};
 453     char mem_name[32];
 454     hwaddr node0_size, mem_start;
 455     uint64_t mem_reg_property[2];
 456     int i, off;
 457
 458     /* memory node(s) */
 459     node0_size = (nb_numa_nodes > 1) ? node_mem[0] : ram_size;
 460     if (spapr->rma_size > node0_size) {
 461         spapr->rma_size = node0_size;
 462     }
 463
 464     /* RMA */
 465     mem_reg_property[0] = 0;
 466     mem_reg_property[1] = cpu_to_be64(spapr->rma_size);
 467     off = fdt_add_subnode(fdt, 0, "memory@0");
 468     _FDT(off);
 469     _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
 470     _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
 471                       sizeof(mem_reg_property))));
 472     _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
 473                       sizeof(associativity))));
 474
 475     /* RAM: Node 0 */
 476     if (node0_size > spapr->rma_size) {
 477         mem_reg_property[0] = cpu_to_be64(spapr->rma_size);
 478         mem_reg_property[1] = cpu_to_be64(node0_size - spapr->rma_size);
 479
 480         sprintf(mem_name, "memory@" TARGET_FMT_lx, spapr->rma_size);
 481         off = fdt_add_subnode(fdt, 0, mem_name);
 482         _FDT(off);
 483         _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
 484         _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
 485                           sizeof(mem_reg_property))));
 486         _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
 487                           sizeof(associativity))));
 488     }
 489
 490     /* RAM: Node 1 and beyond */
 491     mem_start = node0_size;
 492     for (i = 1; i < nb_numa_nodes; i++) {
 493         mem_reg_property[0] = cpu_to_be64(mem_start);
 494         mem_reg_property[1] = cpu_to_be64(node_mem[i]);
 495         associativity[3] = associativity[4] = cpu_to_be32(i);
 496         sprintf(mem_name, "memory@" TARGET_FMT_lx, mem_start);
 497         off = fdt_add_subnode(fdt, 0, mem_name);
 498         _FDT(off);
 499         _FDT((fdt_setprop_string(fdt, off, "device_type", "memory")));
 500         _FDT((fdt_setprop(fdt, off, "reg", mem_reg_property,
 501                           sizeof(mem_reg_property))));
 502         _FDT((fdt_setprop(fdt, off, "ibm,associativity", associativity,
 503                           sizeof(associativity))));
 504         mem_start += node_mem[i];
 505     }
 506
 507     return 0;
 508 }
 509
 510 static void spapr_finalize_fdt(sPAPREnvironment *spapr,
 511                                hwaddr fdt_addr,
 512                                hwaddr rtas_addr,
 513                                hwaddr rtas_size)
 514 {
 515     int ret;
 516     void *fdt;
 517     sPAPRPHBState *phb;
 518
 519     fdt = g_malloc(FDT_MAX_SIZE);
 520
 521     /* open out the base tree into a temp buffer for the final tweaks */
 522     _FDT((fdt_open_into(spapr->fdt_skel, fdt, FDT_MAX_SIZE)));
 523
 524     ret = spapr_populate_memory(spapr, fdt);
 525     if (ret < 0) {
 526         fprintf(stderr, "couldn't setup memory nodes in fdt\n");
 527         exit(1);
 528     }
 529
 530     ret = spapr_populate_vdevice(spapr->vio_bus, fdt);
 531     if (ret < 0) {
 532         fprintf(stderr, "couldn't setup vio devices in fdt\n");
 533         exit(1);
 534     }
 535
 536     QLIST_FOREACH(phb, &spapr->phbs, list) {
 537         ret = spapr_populate_pci_dt(phb, PHANDLE_XICP, fdt);
 538     }
 539
 540     if (ret < 0) {
 541         fprintf(stderr, "couldn't setup PCI devices in fdt\n");
 542         exit(1);
 543     }
 544
 545     /* RTAS */
 546     ret = spapr_rtas_device_tree_setup(fdt, rtas_addr, rtas_size);
 547     if (ret < 0) {
 548         fprintf(stderr, "Couldn't set up RTAS device tree properties\n");
 549     }
 550
 551     /* Advertise NUMA via ibm,associativity */
 552     ret = spapr_fixup_cpu_dt(fdt, spapr);
 553     if (ret < 0) {
 554         fprintf(stderr, "Couldn't finalize CPU device tree properties\n");
 555     }
 556
 557     if (!spapr->has_graphics) {
 558         spapr_populate_chosen_stdout(fdt, spapr->vio_bus);
 559     }
 560
 561     _FDT((fdt_pack(fdt)));
 562
 563     if (fdt_totalsize(fdt) > FDT_MAX_SIZE) {
 564         hw_error("FDT too big ! 0x%x bytes (max is 0x%x)\n",
 565                  fdt_totalsize(fdt), FDT_MAX_SIZE);
 566         exit(1);
 567     }
 568
 569     cpu_physical_memory_write(fdt_addr, fdt, fdt_totalsize(fdt));
 570
 571     g_free(fdt);
 572 }
 573
 574 static uint64_t translate_kernel_address(void *opaque, uint64_t addr)
 575 {
 576     return (addr & 0x0fffffff) + KERNEL_LOAD_ADDR;
 577 }
 578
 579 static void emulate_spapr_hypercall(PowerPCCPU *cpu)
 580 {
 581     CPUPPCState *env = &cpu->env;
 582
 583     if (msr_pr) {
 584         hcall_dprintf("Hypercall made with MSR[PR]=1\n");
 585         env->gpr[3] = H_PRIVILEGE;
 586     } else {
 587         env->gpr[3] = spapr_hypercall(cpu, env->gpr[3], &env->gpr[4]);
 588     }
 589 }
 590
 591 static void spapr_reset_htab(sPAPREnvironment *spapr)
 592 {
 593     long shift;
 594
 595     /* allocate hash page table.  For now we always make this 16mb,
 596      * later we should probably make it scale to the size of guest
 597      * RAM */
 598
 599     shift = kvmppc_reset_htab(spapr->htab_shift);
 600
 601     if (shift > 0) {
 602         /* Kernel handles htab, we don't need to allocate one */
 603         spapr->htab_shift = shift;
 604     } else {
 605         if (!spapr->htab) {
 606             /* Allocate an htab if we don't yet have one */
 607             spapr->htab = qemu_memalign(HTAB_SIZE(spapr), HTAB_SIZE(spapr));
 608         }
 609
 610         /* And clear it */
 611         memset(spapr->htab, 0, HTAB_SIZE(spapr));
 612     }
 613
 614     /* Update the RMA size if necessary */
 615     if (spapr->vrma_adjust) {
 616         spapr->rma_size = kvmppc_rma_size(ram_size, spapr->htab_shift);
 617     }
 618 }
 619
 620 static void ppc_spapr_reset(void)
 621 {
 622     /* Reset the hash table & recalc the RMA */
 623     spapr_reset_htab(spapr);
 624
 625     qemu_devices_reset();
 626
 627     /* Load the fdt */
 628     spapr_finalize_fdt(spapr, spapr->fdt_addr, spapr->rtas_addr,
 629                        spapr->rtas_size);
 630
 631     /* Set up the entry state */
 632     first_cpu->gpr[3] = spapr->fdt_addr;
 633     first_cpu->gpr[5] = 0;
 634     first_cpu->halted = 0;
 635     first_cpu->nip = spapr->entry_point;
 636
 637 }
 638
 639 static void spapr_cpu_reset(void *opaque)
 640 {
 641     PowerPCCPU *cpu = opaque;
 642     CPUPPCState *env = &cpu->env;
 643
 644     cpu_reset(CPU(cpu));
 645
 646     /* All CPUs start halted.  CPU0 is unhalted from the machine level
 647      * reset code and the rest are explicitly started up by the guest
 648      * using an RTAS call */
 649     env->halted = 1;
 650
 651     env->spr[SPR_HIOR] = 0;
 652
 653     env->external_htab = spapr->htab;
 654     env->htab_base = -1;
 655     env->htab_mask = HTAB_SIZE(spapr) - 1;
 656     env->spr[SPR_SDR1] = (unsigned long)spapr->htab |
 657         (spapr->htab_shift - 18);
 658 }
 659
 660 static void spapr_create_nvram(sPAPREnvironment *spapr)
 661 {
 662     QemuOpts *machine_opts;
 663     DeviceState *dev;
 664
 665     dev = qdev_create(&spapr->vio_bus->bus, "spapr-nvram");
 666
 667     machine_opts = qemu_opts_find(qemu_find_opts("machine"), 0);
 668     if (machine_opts) {
 669         const char *drivename;
 670
 671         drivename = qemu_opt_get(machine_opts, "nvram");
 672         if (drivename) {
 673             BlockDriverState *bs;
 674
 675             bs = bdrv_find(drivename);
 676             if (!bs) {
 677                 fprintf(stderr, "No such block device \"%s\" for nvram\n",
 678                         drivename);
 679                 exit(1);
 680             }
 681             qdev_prop_set_drive_nofail(dev, "drive", bs);
 682         }
 683     }
 684
 685     qdev_init_nofail(dev);
 686
 687     spapr->nvram = (struct sPAPRNVRAM *)dev;
 688 }
 689
 690 /* Returns whether we want to use VGA or not */
 691 static int spapr_vga_init(PCIBus *pci_bus)
 692 {
 693     switch (vga_interface_type) {
 694     case VGA_NONE:
 695     case VGA_STD:
 696         return pci_vga_init(pci_bus) != NULL;
 697     default:
 698         fprintf(stderr, "This vga model is not supported,"
 699                 "currently it only supports -vga std\n");
 700         exit(0);
 701         break;
 702     }
 703 }
 704
 705 /* pSeries LPAR / sPAPR hardware init */
 706 static void ppc_spapr_init(QEMUMachineInitArgs *args)
 707 {
 708     ram_addr_t ram_size = args->ram_size;
 709     const char *cpu_model = args->cpu_model;
 710     const char *kernel_filename = args->kernel_filename;
 711     const char *kernel_cmdline = args->kernel_cmdline;
 712     const char *initrd_filename = args->initrd_filename;
 713     const char *boot_device = args->boot_device;
 714     PowerPCCPU *cpu;
 715     CPUPPCState *env;
 716     PCIHostState *phb;
 717     int i;
 718     MemoryRegion *sysmem = get_system_memory();
 719     MemoryRegion *ram = g_new(MemoryRegion, 1);
 720     hwaddr rma_alloc_size;
 721     uint32_t initrd_base = 0;
 722     long kernel_size = 0, initrd_size = 0;
 723     long load_limit, rtas_limit, fw_size;
 724     char *filename;
 725
 726     msi_supported = true;
 727
 728     spapr = g_malloc0(sizeof(*spapr));
 729     QLIST_INIT(&spapr->phbs);
 730
 731     cpu_ppc_hypercall = emulate_spapr_hypercall;
 732
 733     /* Allocate RMA if necessary */
 734     rma_alloc_size = kvmppc_alloc_rma("ppc_spapr.rma", sysmem);
 735
 736     if (rma_alloc_size == -1) {
 737         hw_error("qemu: Unable to create RMA\n");
 738         exit(1);
 739     }
 740
 741     if (rma_alloc_size && (rma_alloc_size < ram_size)) {
 742         spapr->rma_size = rma_alloc_size;
 743     } else {
 744         spapr->rma_size = ram_size;
 745
 746         /* With KVM, we don't actually know whether KVM supports an
 747          * unbounded RMA (PR KVM) or is limited by the hash table size
 748          * (HV KVM using VRMA), so we always assume the latter
 749          *
 750          * In that case, we also limit the initial allocations for RTAS
 751          * etc... to 256M since we have no way to know what the VRMA size
 752          * is going to be as it depends on the size of the hash table
 753          * isn't determined yet.
 754          */
 755         if (kvm_enabled()) {
 756             spapr->vrma_adjust = 1;
 757             spapr->rma_size = MIN(spapr->rma_size, 0x10000000);
 758         }
 759     }
 760
 761     /* We place the device tree and RTAS just below either the top of the RMA,
 762      * or just below 2GB, whichever is lowere, so that it can be
 763      * processed with 32-bit real mode code if necessary */
 764     rtas_limit = MIN(spapr->rma_size, 0x80000000);
 765     spapr->rtas_addr = rtas_limit - RTAS_MAX_SIZE;
 766     spapr->fdt_addr = spapr->rtas_addr - FDT_MAX_SIZE;
 767     load_limit = spapr->fdt_addr - FW_OVERHEAD;
 768
 769     /* We aim for a hash table of size 1/128 the size of RAM.  The
 770      * normal rule of thumb is 1/64 the size of RAM, but that's much
 771      * more than needed for the Linux guests we support. */
 772     spapr->htab_shift = 18; /* Minimum architected size */
 773     while (spapr->htab_shift <= 46) {
 774         if ((1ULL << (spapr->htab_shift + 7)) >= ram_size) {
 775             break;
 776         }
 777         spapr->htab_shift++;
 778     }
 779
 780     /* init CPUs */
 781     if (cpu_model == NULL) {
 782         cpu_model = kvm_enabled() ? "host" : "POWER7";
 783     }
 784     for (i = 0; i < smp_cpus; i++) {
 785         cpu = cpu_ppc_init(cpu_model);
 786         if (cpu == NULL) {
 787             fprintf(stderr, "Unable to find PowerPC CPU definition\n");
 788             exit(1);
 789         }
 790         env = &cpu->env;
 791
 792         /* Set time-base frequency to 512 MHz */
 793         cpu_ppc_tb_init(env, TIMEBASE_FREQ);
 794
 795         /* PAPR always has exception vectors in RAM not ROM */
 796         env->hreset_excp_prefix = 0;
 797
 798         /* Tell KVM that we're in PAPR mode */
 799         if (kvm_enabled()) {
 800             kvmppc_set_papr(env);
 801         }
 802
 803         qemu_register_reset(spapr_cpu_reset, cpu);
 804     }
 805
 806     /* allocate RAM */
 807     spapr->ram_limit = ram_size;
 808     if (spapr->ram_limit > rma_alloc_size) {
 809         ram_addr_t nonrma_base = rma_alloc_size;
 810         ram_addr_t nonrma_size = spapr->ram_limit - rma_alloc_size;
 811
 812         memory_region_init_ram(ram, "ppc_spapr.ram", nonrma_size);
 813         vmstate_register_ram_global(ram);
 814         memory_region_add_subregion(sysmem, nonrma_base, ram);
 815     }
 816
 817     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, "spapr-rtas.bin");
 818     spapr->rtas_size = load_image_targphys(filename, spapr->rtas_addr,
 819                                            rtas_limit - spapr->rtas_addr);
 820     if (spapr->rtas_size < 0) {
 821         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
 822         exit(1);
 823     }
 824     if (spapr->rtas_size > RTAS_MAX_SIZE) {
 825         hw_error("RTAS too big ! 0x%lx bytes (max is 0x%x)\n",
 826                  spapr->rtas_size, RTAS_MAX_SIZE);
 827         exit(1);
 828     }
 829     g_free(filename);
 830
 831
 832     /* Set up Interrupt Controller */
 833     spapr->icp = xics_system_init(XICS_IRQS);
 834     spapr->next_irq = XICS_IRQ_BASE;
 835
 836     /* Set up EPOW events infrastructure */
 837     spapr_events_init(spapr);
 838
 839     /* Set up IOMMU */
 840     spapr_iommu_init();
 841
 842     /* Set up VIO bus */
 843     spapr->vio_bus = spapr_vio_bus_init();
 844
 845     for (i = 0; i < MAX_SERIAL_PORTS; i++) {
 846         if (serial_hds[i]) {
 847             spapr_vty_create(spapr->vio_bus, serial_hds[i]);
 848         }
 849     }
 850
 851     /* We always have at least the nvram device on VIO */
 852     spapr_create_nvram(spapr);
 853
 854     /* Set up PCI */
 855     spapr_pci_rtas_init();
 856
 857     spapr_create_phb(spapr, "pci", SPAPR_PCI_BUID,
 858                      SPAPR_PCI_MEM_WIN_ADDR,
 859                      SPAPR_PCI_MEM_WIN_SIZE,
 860                      SPAPR_PCI_IO_WIN_ADDR,
 861                      SPAPR_PCI_MSI_WIN_ADDR);
 862     phb = PCI_HOST_BRIDGE(QLIST_FIRST(&spapr->phbs));
 863
 864     for (i = 0; i < nb_nics; i++) {
 865         NICInfo *nd = &nd_table[i];
 866
 867         if (!nd->model) {
 868             nd->model = g_strdup("ibmveth");
 869         }
 870
 871         if (strcmp(nd->model, "ibmveth") == 0) {
 872             spapr_vlan_create(spapr->vio_bus, nd);
 873         } else {
 874             pci_nic_init_nofail(&nd_table[i], nd->model, NULL);
 875         }
 876     }
 877
 878     for (i = 0; i <= drive_get_max_bus(IF_SCSI); i++) {
 879         spapr_vscsi_create(spapr->vio_bus);
 880     }
 881
 882     /* Graphics */
 883     if (spapr_vga_init(phb->bus)) {
 884         spapr->has_graphics = true;
 885     }
 886
 887     if (usb_enabled(spapr->has_graphics)) {
 888         pci_create_simple(phb->bus, -1, "pci-ohci");
 889         if (spapr->has_graphics) {
 890             usbdevice_create("keyboard");
 891             usbdevice_create("mouse");
 892         }
 893     }
 894
 895     if (spapr->rma_size < (MIN_RMA_SLOF << 20)) {
 896         fprintf(stderr, "qemu: pSeries SLOF firmware requires >= "
 897                 "%ldM guest RMA (Real Mode Area memory)\n", MIN_RMA_SLOF);
 898         exit(1);
 899     }
 900
 901     if (kernel_filename) {
 902         uint64_t lowaddr = 0;
 903
 904         kernel_size = load_elf(kernel_filename, translate_kernel_address, NULL,
 905                                NULL, &lowaddr, NULL, 1, ELF_MACHINE, 0);
 906         if (kernel_size < 0) {
 907             kernel_size = load_image_targphys(kernel_filename,
 908                                               KERNEL_LOAD_ADDR,
 909                                               load_limit - KERNEL_LOAD_ADDR);
 910         }
 911         if (kernel_size < 0) {
 912             fprintf(stderr, "qemu: could not load kernel '%s'\n",
 913                     kernel_filename);
 914             exit(1);
 915         }
 916
 917         /* load initrd */
 918         if (initrd_filename) {
 919             /* Try to locate the initrd in the gap between the kernel
 920              * and the firmware. Add a bit of space just in case
 921              */
 922             initrd_base = (KERNEL_LOAD_ADDR + kernel_size + 0x1ffff) & ~0xffff;
 923             initrd_size = load_image_targphys(initrd_filename, initrd_base,
 924                                               load_limit - initrd_base);
 925             if (initrd_size < 0) {
 926                 fprintf(stderr, "qemu: could not load initial ram disk '%s'\n",
 927                         initrd_filename);
 928                 exit(1);
 929             }
 930         } else {
 931             initrd_base = 0;
 932             initrd_size = 0;
 933         }
 934     }
 935
 936     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, FW_FILE_NAME);
 937     fw_size = load_image_targphys(filename, 0, FW_MAX_SIZE);
 938     if (fw_size < 0) {
 939         hw_error("qemu: could not load LPAR rtas '%s'\n", filename);
 940         exit(1);
 941     }
 942     g_free(filename);
 943
 944     spapr->entry_point = 0x100;
 945
 946     /* Prepare the device tree */
 947     spapr->fdt_skel = spapr_create_fdt_skel(cpu_model,
 948                                             initrd_base, initrd_size,
 949                                             kernel_size,
 950                                             boot_device, kernel_cmdline,
 951                                             spapr->epow_irq);
 952     assert(spapr->fdt_skel != NULL);
 953 }
 954
 955 static QEMUMachine spapr_machine = {
 956     .name = "pseries",
 957     .desc = "pSeries Logical Partition (PAPR compliant)",
 958     .init = ppc_spapr_init,
 959     .reset = ppc_spapr_reset,
 960     .block_default_type = IF_SCSI,
 961     .max_cpus = MAX_CPUS,
 962     .no_parallel = 1,
 963 };
 964
 965 static void spapr_machine_init(void)
 966 {
 967     qemu_register_machine(&spapr_machine);
 968 }
 969
 970 machine_init(spapr_machine_init);