arch/s390/kernel/setup.c

   1 /*
   2  *  arch/s390/kernel/setup.c
   3  *
   4  *  S390 version
   5  *    Copyright (C) IBM Corp. 1999,2010
   6  *    Author(s): Hartmut Penner (hp@de.ibm.com),
   7  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   8  *
   9  *  Derived from "arch/i386/kernel/setup.c"
  10  *    Copyright (C) 1995, Linus Torvalds
  11  */
  12
  13 /*
  14  * This file handles the architecture-dependent parts of initialization
  15  */
  16
  17 #define KMSG_COMPONENT "setup"
  18 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  19
  20 #include <linux/errno.h>
  21 #include <linux/module.h>
  22 #include <linux/sched.h>
  23 #include <linux/kernel.h>
  24 #include <linux/mm.h>
  25 #include <linux/stddef.h>
  26 #include <linux/unistd.h>
  27 #include <linux/ptrace.h>
  28 #include <linux/user.h>
  29 #include <linux/tty.h>
  30 #include <linux/ioport.h>
  31 #include <linux/delay.h>
  32 #include <linux/init.h>
  33 #include <linux/initrd.h>
  34 #include <linux/bootmem.h>
  35 #include <linux/root_dev.h>
  36 #include <linux/console.h>
  37 #include <linux/kernel_stat.h>
  38 #include <linux/device.h>
  39 #include <linux/notifier.h>
  40 #include <linux/pfn.h>
  41 #include <linux/ctype.h>
  42 #include <linux/reboot.h>
  43 #include <linux/topology.h>
  44 #include <linux/ftrace.h>
  45
  46 #include <asm/ipl.h>
  47 #include <asm/uaccess.h>
  48 #include <asm/system.h>
  49 #include <asm/smp.h>
  50 #include <asm/mmu_context.h>
  51 #include <asm/cpcmd.h>
  52 #include <asm/lowcore.h>
  53 #include <asm/irq.h>
  54 #include <asm/page.h>
  55 #include <asm/ptrace.h>
  56 #include <asm/sections.h>
  57 #include <asm/ebcdic.h>
  58 #include <asm/compat.h>
  59 #include <asm/kvm_virtio.h>
  60
  61 long psw_kernel_bits    = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_PRIMARY |
  62                            PSW_MASK_MCHECK | PSW_DEFAULT_KEY);
  63 long psw_user_bits      = (PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
  64                            PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
  65                            PSW_MASK_PSTATE | PSW_DEFAULT_KEY);
  66
  67 /*
  68  * User copy operations.
  69  */
  70 struct uaccess_ops uaccess;
  71 EXPORT_SYMBOL(uaccess);
  72
  73 /*
  74  * Machine setup..
  75  */
  76 unsigned int console_mode = 0;
  77 EXPORT_SYMBOL(console_mode);
  78
  79 unsigned int console_devno = -1;
  80 EXPORT_SYMBOL(console_devno);
  81
  82 unsigned int console_irq = -1;
  83 EXPORT_SYMBOL(console_irq);
  84
  85 unsigned long elf_hwcap = 0;
  86 char elf_platform[ELF_PLATFORM_SIZE];
  87
  88 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
  89
  90 int __initdata memory_end_set;
  91 unsigned long __initdata memory_end;
  92
  93 /* An array with a pointer to the lowcore of every CPU. */
  94 struct _lowcore *lowcore_ptr[NR_CPUS];
  95 EXPORT_SYMBOL(lowcore_ptr);
  96
  97 /*
  98  * This is set up by the setup-routine at boot-time
  99  * for S390 need to find out, what we have to setup
 100  * using address 0x10400 ...
 101  */
 102
 103 #include <asm/setup.h>
 104
 105 static struct resource code_resource = {
 106         .name  = "Kernel code",
 107         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 108 };
 109
 110 static struct resource data_resource = {
 111         .name = "Kernel data",
 112         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 113 };
 114
 115 /*
 116  * condev= and conmode= setup parameter.
 117  */
 118
 119 static int __init condev_setup(char *str)
 120 {
 121         int vdev;
 122
 123         vdev = simple_strtoul(str, &str, 0);
 124         if (vdev >= 0 && vdev < 65536) {
 125                 console_devno = vdev;
 126                 console_irq = -1;
 127         }
 128         return 1;
 129 }
 130
 131 __setup("condev=", condev_setup);
 132
 133 static void __init set_preferred_console(void)
 134 {
 135         if (MACHINE_IS_KVM)
 136                 add_preferred_console("hvc", 0, NULL);
 137         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 138                 add_preferred_console("ttyS", 0, NULL);
 139         else if (CONSOLE_IS_3270)
 140                 add_preferred_console("tty3270", 0, NULL);
 141 }
 142
 143 static int __init conmode_setup(char *str)
 144 {
 145 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 146         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 147                 SET_CONSOLE_SCLP;
 148 #endif
 149 #if defined(CONFIG_TN3215_CONSOLE)
 150         if (strncmp(str, "3215", 5) == 0)
 151                 SET_CONSOLE_3215;
 152 #endif
 153 #if defined(CONFIG_TN3270_CONSOLE)
 154         if (strncmp(str, "3270", 5) == 0)
 155                 SET_CONSOLE_3270;
 156 #endif
 157         set_preferred_console();
 158         return 1;
 159 }
 160
 161 __setup("conmode=", conmode_setup);
 162
 163 static void __init conmode_default(void)
 164 {
 165         char query_buffer[1024];
 166         char *ptr;
 167
 168         if (MACHINE_IS_VM) {
 169                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 170                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 171                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 172                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 173                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 174                 ptr = strstr(query_buffer, "CONMODE");
 175                 /*
 176                  * Set the conmode to 3215 so that the device recognition
 177                  * will set the cu_type of the console to 3215. If the
 178                  * conmode is 3270 and we don't set it back then both
 179                  * 3215 and the 3270 driver will try to access the console
 180                  * device (3215 as console and 3270 as normal tty).
 181                  */
 182                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 183                 if (ptr == NULL) {
 184 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 185                         SET_CONSOLE_SCLP;
 186 #endif
 187                         return;
 188                 }
 189                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 190 #if defined(CONFIG_TN3270_CONSOLE)
 191                         SET_CONSOLE_3270;
 192 #elif defined(CONFIG_TN3215_CONSOLE)
 193                         SET_CONSOLE_3215;
 194 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 195                         SET_CONSOLE_SCLP;
 196 #endif
 197                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 198 #if defined(CONFIG_TN3215_CONSOLE)
 199                         SET_CONSOLE_3215;
 200 #elif defined(CONFIG_TN3270_CONSOLE)
 201                         SET_CONSOLE_3270;
 202 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 203                         SET_CONSOLE_SCLP;
 204 #endif
 205                 }
 206         } else {
 207 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 208                 SET_CONSOLE_SCLP;
 209 #endif
 210         }
 211 }
 212
 213 #ifdef CONFIG_ZFCPDUMP
 214 static void __init setup_zfcpdump(unsigned int console_devno)
 215 {
 216         static char str[41];
 217
 218         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 219                 return;
 220         if (console_devno != -1)
 221                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 222                         ipl_info.data.fcp.dev_id.devno, console_devno);
 223         else
 224                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 225                         ipl_info.data.fcp.dev_id.devno);
 226         strcat(boot_command_line, str);
 227         console_loglevel = 2;
 228 }
 229 #else
 230 static inline void setup_zfcpdump(unsigned int console_devno) {}
 231 #endif /* CONFIG_ZFCPDUMP */
 232
 233  /*
 234  * Reboot, halt and power_off stubs. They just call _machine_restart,
 235  * _machine_halt or _machine_power_off.
 236  */
 237
 238 void machine_restart(char *command)
 239 {
 240         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 241                 /*
 242                  * Only unblank the console if we are called in enabled
 243                  * context or a bust_spinlocks cleared the way for us.
 244                  */
 245                 console_unblank();
 246         _machine_restart(command);
 247 }
 248
 249 void machine_halt(void)
 250 {
 251         if (!in_interrupt() || oops_in_progress)
 252                 /*
 253                  * Only unblank the console if we are called in enabled
 254                  * context or a bust_spinlocks cleared the way for us.
 255                  */
 256                 console_unblank();
 257         _machine_halt();
 258 }
 259
 260 void machine_power_off(void)
 261 {
 262         if (!in_interrupt() || oops_in_progress)
 263                 /*
 264                  * Only unblank the console if we are called in enabled
 265                  * context or a bust_spinlocks cleared the way for us.
 266                  */
 267                 console_unblank();
 268         _machine_power_off();
 269 }
 270
 271 /*
 272  * Dummy power off function.
 273  */
 274 void (*pm_power_off)(void) = machine_power_off;
 275
 276 static int __init early_parse_mem(char *p)
 277 {
 278         memory_end = memparse(p, &p);
 279         memory_end_set = 1;
 280         return 0;
 281 }
 282 early_param("mem", early_parse_mem);
 283
 284 unsigned int user_mode = HOME_SPACE_MODE;
 285 EXPORT_SYMBOL_GPL(user_mode);
 286
 287 static int set_amode_and_uaccess(unsigned long user_amode,
 288                                  unsigned long user32_amode)
 289 {
 290         psw_user_bits = PSW_BASE_BITS | PSW_MASK_DAT | user_amode |
 291                         PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
 292                         PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
 293 #ifdef CONFIG_COMPAT
 294         psw_user32_bits = PSW_BASE32_BITS | PSW_MASK_DAT | user_amode |
 295                           PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK |
 296                           PSW_MASK_PSTATE | PSW_DEFAULT_KEY;
 297         psw32_user_bits = PSW32_BASE_BITS | PSW32_MASK_DAT | user32_amode |
 298                           PSW32_MASK_IO | PSW32_MASK_EXT | PSW32_MASK_MCHECK |
 299                           PSW32_MASK_PSTATE;
 300 #endif
 301         psw_kernel_bits = PSW_BASE_BITS | PSW_MASK_DAT | PSW_ASC_HOME |
 302                           PSW_MASK_MCHECK | PSW_DEFAULT_KEY;
 303
 304         if (MACHINE_HAS_MVCOS) {
 305                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
 306                 return 1;
 307         } else {
 308                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
 309                 return 0;
 310         }
 311 }
 312
 313 /*
 314  * Switch kernel/user addressing modes?
 315  */
 316 static int __init early_parse_switch_amode(char *p)
 317 {
 318         if (user_mode != SECONDARY_SPACE_MODE)
 319                 user_mode = PRIMARY_SPACE_MODE;
 320         return 0;
 321 }
 322 early_param("switch_amode", early_parse_switch_amode);
 323
 324 static int __init early_parse_user_mode(char *p)
 325 {
 326         if (p && strcmp(p, "primary") == 0)
 327                 user_mode = PRIMARY_SPACE_MODE;
 328 #ifdef CONFIG_S390_EXEC_PROTECT
 329         else if (p && strcmp(p, "secondary") == 0)
 330                 user_mode = SECONDARY_SPACE_MODE;
 331 #endif
 332         else if (!p || strcmp(p, "home") == 0)
 333                 user_mode = HOME_SPACE_MODE;
 334         else
 335                 return 1;
 336         return 0;
 337 }
 338 early_param("user_mode", early_parse_user_mode);
 339
 340 #ifdef CONFIG_S390_EXEC_PROTECT
 341 /*
 342  * Enable execute protection?
 343  */
 344 static int __init early_parse_noexec(char *p)
 345 {
 346         if (!strncmp(p, "off", 3))
 347                 return 0;
 348         user_mode = SECONDARY_SPACE_MODE;
 349         return 0;
 350 }
 351 early_param("noexec", early_parse_noexec);
 352 #endif /* CONFIG_S390_EXEC_PROTECT */
 353
 354 static void setup_addressing_mode(void)
 355 {
 356         if (user_mode == SECONDARY_SPACE_MODE) {
 357                 if (set_amode_and_uaccess(PSW_ASC_SECONDARY,
 358                                           PSW32_ASC_SECONDARY))
 359                         pr_info("Execute protection active, "
 360                                 "mvcos available\n");
 361                 else
 362                         pr_info("Execute protection active, "
 363                                 "mvcos not available\n");
 364         } else if (user_mode == PRIMARY_SPACE_MODE) {
 365                 if (set_amode_and_uaccess(PSW_ASC_PRIMARY, PSW32_ASC_PRIMARY))
 366                         pr_info("Address spaces switched, "
 367                                 "mvcos available\n");
 368                 else
 369                         pr_info("Address spaces switched, "
 370                                 "mvcos not available\n");
 371         }
 372 }
 373
 374 static void __init
 375 setup_lowcore(void)
 376 {
 377         struct _lowcore *lc;
 378
 379         /*
 380          * Setup lowcore for boot cpu
 381          */
 382         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 383         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 384         lc->restart_psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY;
 385         lc->restart_psw.addr =
 386                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 387         if (user_mode != HOME_SPACE_MODE)
 388                 lc->restart_psw.mask |= PSW_ASC_HOME;
 389         lc->external_new_psw.mask = psw_kernel_bits;
 390         lc->external_new_psw.addr =
 391                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 392         lc->svc_new_psw.mask = psw_kernel_bits | PSW_MASK_IO | PSW_MASK_EXT;
 393         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 394         lc->program_new_psw.mask = psw_kernel_bits;
 395         lc->program_new_psw.addr =
 396                 PSW_ADDR_AMODE | (unsigned long)pgm_check_handler;
 397         lc->mcck_new_psw.mask =
 398                 psw_kernel_bits & ~PSW_MASK_MCHECK & ~PSW_MASK_DAT;
 399         lc->mcck_new_psw.addr =
 400                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 401         lc->io_new_psw.mask = psw_kernel_bits;
 402         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 403         lc->clock_comparator = -1ULL;
 404         lc->cmf_hpp = -1ULL;
 405         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 406         lc->async_stack = (unsigned long)
 407                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 408         lc->panic_stack = (unsigned long)
 409                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 410         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 411         lc->thread_info = (unsigned long) &init_thread_union;
 412         lc->machine_flags = S390_lowcore.machine_flags;
 413 #ifndef CONFIG_64BIT
 414         if (MACHINE_HAS_IEEE) {
 415                 lc->extended_save_area_addr = (__u32)
 416                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 417                 /* enable extended save area */
 418                 __ctl_set_bit(14, 29);
 419         }
 420 #else
 421         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 422 #endif
 423         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 424         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 425         lc->exit_timer = S390_lowcore.exit_timer;
 426         lc->user_timer = S390_lowcore.user_timer;
 427         lc->system_timer = S390_lowcore.system_timer;
 428         lc->steal_timer = S390_lowcore.steal_timer;
 429         lc->last_update_timer = S390_lowcore.last_update_timer;
 430         lc->last_update_clock = S390_lowcore.last_update_clock;
 431         lc->ftrace_func = S390_lowcore.ftrace_func;
 432         set_prefix((u32)(unsigned long) lc);
 433         lowcore_ptr[0] = lc;
 434 }
 435
 436 static void __init
 437 setup_resources(void)
 438 {
 439         struct resource *res, *sub_res;
 440         int i;
 441
 442         code_resource.start = (unsigned long) &_text;
 443         code_resource.end = (unsigned long) &_etext - 1;
 444         data_resource.start = (unsigned long) &_etext;
 445         data_resource.end = (unsigned long) &_edata - 1;
 446
 447         for (i = 0; i < MEMORY_CHUNKS; i++) {
 448                 if (!memory_chunk[i].size)
 449                         continue;
 450                 res = alloc_bootmem_low(sizeof(struct resource));
 451                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 452                 switch (memory_chunk[i].type) {
 453                 case CHUNK_READ_WRITE:
 454                         res->name = "System RAM";
 455                         break;
 456                 case CHUNK_READ_ONLY:
 457                         res->name = "System ROM";
 458                         res->flags |= IORESOURCE_READONLY;
 459                         break;
 460                 default:
 461                         res->name = "reserved";
 462                 }
 463                 res->start = memory_chunk[i].addr;
 464                 res->end = memory_chunk[i].addr +  memory_chunk[i].size - 1;
 465                 request_resource(&iomem_resource, res);
 466
 467                 if (code_resource.start >= res->start  &&
 468                         code_resource.start <= res->end &&
 469                         code_resource.end > res->end) {
 470                         sub_res = alloc_bootmem_low(sizeof(struct resource));
 471                         memcpy(sub_res, &code_resource,
 472                                 sizeof(struct resource));
 473                         sub_res->end = res->end;
 474                         code_resource.start = res->end + 1;
 475                         request_resource(res, sub_res);
 476                 }
 477
 478                 if (code_resource.start >= res->start &&
 479                         code_resource.start <= res->end &&
 480                         code_resource.end <= res->end)
 481                         request_resource(res, &code_resource);
 482
 483                 if (data_resource.start >= res->start &&
 484                         data_resource.start <= res->end &&
 485                         data_resource.end > res->end) {
 486                         sub_res = alloc_bootmem_low(sizeof(struct resource));
 487                         memcpy(sub_res, &data_resource,
 488                                 sizeof(struct resource));
 489                         sub_res->end = res->end;
 490                         data_resource.start = res->end + 1;
 491                         request_resource(res, sub_res);
 492                 }
 493
 494                 if (data_resource.start >= res->start &&
 495                         data_resource.start <= res->end &&
 496                         data_resource.end <= res->end)
 497                         request_resource(res, &data_resource);
 498         }
 499 }
 500
 501 unsigned long real_memory_size;
 502 EXPORT_SYMBOL_GPL(real_memory_size);
 503
 504 static void __init setup_memory_end(void)
 505 {
 506         unsigned long memory_size;
 507         unsigned long max_mem;
 508         int i;
 509
 510 #ifdef CONFIG_ZFCPDUMP
 511         if (ipl_info.type == IPL_TYPE_FCP_DUMP) {
 512                 memory_end = ZFCPDUMP_HSA_SIZE;
 513                 memory_end_set = 1;
 514         }
 515 #endif
 516         memory_size = 0;
 517         memory_end &= PAGE_MASK;
 518
 519         max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
 520         memory_end = min(max_mem, memory_end);
 521
 522         /*
 523          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 524          * extra checks that HOLES_IN_ZONE would require.
 525          */
 526         for (i = 0; i < MEMORY_CHUNKS; i++) {
 527                 unsigned long start, end;
 528                 struct mem_chunk *chunk;
 529                 unsigned long align;
 530
 531                 chunk = &memory_chunk[i];
 532                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 533                 start = (chunk->addr + align - 1) & ~(align - 1);
 534                 end = (chunk->addr + chunk->size) & ~(align - 1);
 535                 if (start >= end)
 536                         memset(chunk, 0, sizeof(*chunk));
 537                 else {
 538                         chunk->addr = start;
 539                         chunk->size = end - start;
 540                 }
 541         }
 542
 543         for (i = 0; i < MEMORY_CHUNKS; i++) {
 544                 struct mem_chunk *chunk = &memory_chunk[i];
 545
 546                 real_memory_size = max(real_memory_size,
 547                                        chunk->addr + chunk->size);
 548                 if (chunk->addr >= max_mem) {
 549                         memset(chunk, 0, sizeof(*chunk));
 550                         continue;
 551                 }
 552                 if (chunk->addr + chunk->size > max_mem)
 553                         chunk->size = max_mem - chunk->addr;
 554                 memory_size = max(memory_size, chunk->addr + chunk->size);
 555         }
 556         if (!memory_end)
 557                 memory_end = memory_size;
 558 }
 559
 560 static void __init
 561 setup_memory(void)
 562 {
 563         unsigned long bootmap_size;
 564         unsigned long start_pfn, end_pfn;
 565         int i;
 566
 567         /*
 568          * partially used pages are not usable - thus
 569          * we are rounding upwards:
 570          */
 571         start_pfn = PFN_UP(__pa(&_end));
 572         end_pfn = max_pfn = PFN_DOWN(memory_end);
 573
 574 #ifdef CONFIG_BLK_DEV_INITRD
 575         /*
 576          * Move the initrd in case the bitmap of the bootmem allocater
 577          * would overwrite it.
 578          */
 579
 580         if (INITRD_START && INITRD_SIZE) {
 581                 unsigned long bmap_size;
 582                 unsigned long start;
 583
 584                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 585                 bmap_size = PFN_PHYS(bmap_size);
 586
 587                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 588                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 589
 590                         if (start + INITRD_SIZE > memory_end) {
 591                                 pr_err("initrd extends beyond end of "
 592                                        "memory (0x%08lx > 0x%08lx) "
 593                                        "disabling initrd\n",
 594                                        start + INITRD_SIZE, memory_end);
 595                                 INITRD_START = INITRD_SIZE = 0;
 596                         } else {
 597                                 pr_info("Moving initrd (0x%08lx -> "
 598                                         "0x%08lx, size: %ld)\n",
 599                                         INITRD_START, start, INITRD_SIZE);
 600                                 memmove((void *) start, (void *) INITRD_START,
 601                                         INITRD_SIZE);
 602                                 INITRD_START = start;
 603                         }
 604                 }
 605         }
 606 #endif
 607
 608         /*
 609          * Initialize the boot-time allocator
 610          */
 611         bootmap_size = init_bootmem(start_pfn, end_pfn);
 612
 613         /*
 614          * Register RAM areas with the bootmem allocator.
 615          */
 616
 617         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 618                 unsigned long start_chunk, end_chunk, pfn;
 619
 620                 if (memory_chunk[i].type != CHUNK_READ_WRITE)
 621                         continue;
 622                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 623                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 624                 end_chunk = min(end_chunk, end_pfn);
 625                 if (start_chunk >= end_chunk)
 626                         continue;
 627                 add_active_range(0, start_chunk, end_chunk);
 628                 pfn = max(start_chunk, start_pfn);
 629                 for (; pfn < end_chunk; pfn++)
 630                         page_set_storage_key(PFN_PHYS(pfn), PAGE_DEFAULT_KEY);
 631         }
 632
 633         psw_set_key(PAGE_DEFAULT_KEY);
 634
 635         free_bootmem_with_active_regions(0, max_pfn);
 636
 637         /*
 638          * Reserve memory used for lowcore/command line/kernel image.
 639          */
 640         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 641         reserve_bootmem((unsigned long)_stext,
 642                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 643                         BOOTMEM_DEFAULT);
 644         /*
 645          * Reserve the bootmem bitmap itself as well. We do this in two
 646          * steps (first step was init_bootmem()) because this catches
 647          * the (very unlikely) case of us accidentally initializing the
 648          * bootmem allocator with an invalid RAM area.
 649          */
 650         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 651                         BOOTMEM_DEFAULT);
 652
 653 #ifdef CONFIG_BLK_DEV_INITRD
 654         if (INITRD_START && INITRD_SIZE) {
 655                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 656                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 657                                         BOOTMEM_DEFAULT);
 658                         initrd_start = INITRD_START;
 659                         initrd_end = initrd_start + INITRD_SIZE;
 660                 } else {
 661                         pr_err("initrd extends beyond end of "
 662                                "memory (0x%08lx > 0x%08lx) "
 663                                "disabling initrd\n",
 664                                initrd_start + INITRD_SIZE, memory_end);
 665                         initrd_start = initrd_end = 0;
 666                 }
 667         }
 668 #endif
 669 }
 670
 671 /*
 672  * Setup hardware capabilities.
 673  */
 674 static void __init setup_hwcaps(void)
 675 {
 676         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 677         unsigned long long facility_list_extended;
 678         unsigned int facility_list;
 679         struct cpuid cpu_id;
 680         int i;
 681
 682         facility_list = stfl();
 683         /*
 684          * The store facility list bits numbers as found in the principles
 685          * of operation are numbered with bit 1UL<<31 as number 0 to
 686          * bit 1UL<<0 as number 31.
 687          *   Bit 0: instructions named N3, "backported" to esa-mode
 688          *   Bit 2: z/Architecture mode is active
 689          *   Bit 7: the store-facility-list-extended facility is installed
 690          *   Bit 17: the message-security assist is installed
 691          *   Bit 19: the long-displacement facility is installed
 692          *   Bit 21: the extended-immediate facility is installed
 693          *   Bit 22: extended-translation facility 3 is installed
 694          *   Bit 30: extended-translation facility 3 enhancement facility
 695          * These get translated to:
 696          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 697          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 698          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 699          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 700          */
 701         for (i = 0; i < 6; i++)
 702                 if (facility_list & (1UL << (31 - stfl_bits[i])))
 703                         elf_hwcap |= 1UL << i;
 704
 705         if ((facility_list & (1UL << (31 - 22)))
 706             && (facility_list & (1UL << (31 - 30))))
 707                 elf_hwcap |= HWCAP_S390_ETF3EH;
 708
 709         /*
 710          * Check for additional facilities with store-facility-list-extended.
 711          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 712          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 713          * as stored by stfl, bits 32-xxx contain additional facilities.
 714          * How many facility words are stored depends on the number of
 715          * doublewords passed to the instruction. The additional facilites
 716          * are:
 717          *   Bit 42: decimal floating point facility is installed
 718          *   Bit 44: perform floating point operation facility is installed
 719          * translated to:
 720          *   HWCAP_S390_DFP bit 6 (42 && 44).
 721          */
 722         if ((elf_hwcap & (1UL << 2)) &&
 723             __stfle(&facility_list_extended, 1) > 0) {
 724                 if ((facility_list_extended & (1ULL << (63 - 42)))
 725                     && (facility_list_extended & (1ULL << (63 - 44))))
 726                         elf_hwcap |= HWCAP_S390_DFP;
 727         }
 728
 729         /*
 730          * Huge page support HWCAP_S390_HPAGE is bit 7.
 731          */
 732         if (MACHINE_HAS_HPAGE)
 733                 elf_hwcap |= HWCAP_S390_HPAGE;
 734
 735         /*
 736          * 64-bit register support for 31-bit processes
 737          * HWCAP_S390_HIGH_GPRS is bit 9.
 738          */
 739         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 740
 741         get_cpu_id(&cpu_id);
 742         switch (cpu_id.machine) {
 743         case 0x9672:
 744 #if !defined(CONFIG_64BIT)
 745         default:        /* Use "g5" as default for 31 bit kernels. */
 746 #endif
 747                 strcpy(elf_platform, "g5");
 748                 break;
 749         case 0x2064:
 750         case 0x2066:
 751 #if defined(CONFIG_64BIT)
 752         default:        /* Use "z900" as default for 64 bit kernels. */
 753 #endif
 754                 strcpy(elf_platform, "z900");
 755                 break;
 756         case 0x2084:
 757         case 0x2086:
 758                 strcpy(elf_platform, "z990");
 759                 break;
 760         case 0x2094:
 761         case 0x2096:
 762                 strcpy(elf_platform, "z9-109");
 763                 break;
 764         case 0x2097:
 765         case 0x2098:
 766                 strcpy(elf_platform, "z10");
 767                 break;
 768         }
 769 }
 770
 771 /*
 772  * Setup function called from init/main.c just after the banner
 773  * was printed.
 774  */
 775
 776 void __init
 777 setup_arch(char **cmdline_p)
 778 {
 779         /*
 780          * print what head.S has found out about the machine
 781          */
 782 #ifndef CONFIG_64BIT
 783         if (MACHINE_IS_VM)
 784                 pr_info("Linux is running as a z/VM "
 785                         "guest operating system in 31-bit mode\n");
 786         else if (MACHINE_IS_LPAR)
 787                 pr_info("Linux is running natively in 31-bit mode\n");
 788         if (MACHINE_HAS_IEEE)
 789                 pr_info("The hardware system has IEEE compatible "
 790                         "floating point units\n");
 791         else
 792                 pr_info("The hardware system has no IEEE compatible "
 793                         "floating point units\n");
 794 #else /* CONFIG_64BIT */
 795         if (MACHINE_IS_VM)
 796                 pr_info("Linux is running as a z/VM "
 797                         "guest operating system in 64-bit mode\n");
 798         else if (MACHINE_IS_KVM)
 799                 pr_info("Linux is running under KVM in 64-bit mode\n");
 800         else if (MACHINE_IS_LPAR)
 801                 pr_info("Linux is running natively in 64-bit mode\n");
 802 #endif /* CONFIG_64BIT */
 803
 804         /* Have one command line that is parsed and saved in /proc/cmdline */
 805         /* boot_command_line has been already set up in early.c */
 806         *cmdline_p = boot_command_line;
 807
 808         ROOT_DEV = Root_RAM0;
 809
 810         init_mm.start_code = PAGE_OFFSET;
 811         init_mm.end_code = (unsigned long) &_etext;
 812         init_mm.end_data = (unsigned long) &_edata;
 813         init_mm.brk = (unsigned long) &_end;
 814
 815         if (MACHINE_HAS_MVCOS)
 816                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
 817         else
 818                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
 819
 820         parse_early_param();
 821
 822         setup_ipl();
 823         setup_memory_end();
 824         setup_addressing_mode();
 825         setup_memory();
 826         setup_resources();
 827         setup_lowcore();
 828
 829         cpu_init();
 830         s390_init_cpu_topology();
 831
 832         /*
 833          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
 834          */
 835         setup_hwcaps();
 836
 837         /*
 838          * Create kernel page tables and switch to virtual addressing.
 839          */
 840         paging_init();
 841
 842         /* Setup default console */
 843         conmode_default();
 844         set_preferred_console();
 845
 846         /* Setup zfcpdump support */
 847         setup_zfcpdump(console_devno);
 848 }