Pileus Git - ~andy/linux/blob - 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 #include <linux/kexec.h>
  46 #include <linux/crash_dump.h>
  47 #include <linux/memory.h>
  48
  49 #include <asm/ipl.h>
  50 #include <asm/uaccess.h>
  51 #include <asm/system.h>
  52 #include <asm/smp.h>
  53 #include <asm/mmu_context.h>
  54 #include <asm/cpcmd.h>
  55 #include <asm/lowcore.h>
  56 #include <asm/irq.h>
  57 #include <asm/page.h>
  58 #include <asm/ptrace.h>
  59 #include <asm/sections.h>
  60 #include <asm/ebcdic.h>
  61 #include <asm/compat.h>
  62 #include <asm/kvm_virtio.h>
  63 #include <asm/diag.h>
  64
  65 long psw_kernel_bits    = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
  66                           PSW_MASK_EA | PSW_MASK_BA;
  67 long psw_user_bits      = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
  68                           PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
  69                           PSW_MASK_PSTATE | PSW_ASC_HOME;
  70
  71 /*
  72  * User copy operations.
  73  */
  74 struct uaccess_ops uaccess;
  75 EXPORT_SYMBOL(uaccess);
  76
  77 /*
  78  * Machine setup..
  79  */
  80 unsigned int console_mode = 0;
  81 EXPORT_SYMBOL(console_mode);
  82
  83 unsigned int console_devno = -1;
  84 EXPORT_SYMBOL(console_devno);
  85
  86 unsigned int console_irq = -1;
  87 EXPORT_SYMBOL(console_irq);
  88
  89 unsigned long elf_hwcap = 0;
  90 char elf_platform[ELF_PLATFORM_SIZE];
  91
  92 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
  93
  94 int __initdata memory_end_set;
  95 unsigned long __initdata memory_end;
  96
  97 /* An array with a pointer to the lowcore of every CPU. */
  98 struct _lowcore *lowcore_ptr[NR_CPUS];
  99 EXPORT_SYMBOL(lowcore_ptr);
 100
 101 /*
 102  * This is set up by the setup-routine at boot-time
 103  * for S390 need to find out, what we have to setup
 104  * using address 0x10400 ...
 105  */
 106
 107 #include <asm/setup.h>
 108
 109 /*
 110  * condev= and conmode= setup parameter.
 111  */
 112
 113 static int __init condev_setup(char *str)
 114 {
 115         int vdev;
 116
 117         vdev = simple_strtoul(str, &str, 0);
 118         if (vdev >= 0 && vdev < 65536) {
 119                 console_devno = vdev;
 120                 console_irq = -1;
 121         }
 122         return 1;
 123 }
 124
 125 __setup("condev=", condev_setup);
 126
 127 static void __init set_preferred_console(void)
 128 {
 129         if (MACHINE_IS_KVM)
 130                 add_preferred_console("hvc", 0, NULL);
 131         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 132                 add_preferred_console("ttyS", 0, NULL);
 133         else if (CONSOLE_IS_3270)
 134                 add_preferred_console("tty3270", 0, NULL);
 135 }
 136
 137 static int __init conmode_setup(char *str)
 138 {
 139 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 140         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 141                 SET_CONSOLE_SCLP;
 142 #endif
 143 #if defined(CONFIG_TN3215_CONSOLE)
 144         if (strncmp(str, "3215", 5) == 0)
 145                 SET_CONSOLE_3215;
 146 #endif
 147 #if defined(CONFIG_TN3270_CONSOLE)
 148         if (strncmp(str, "3270", 5) == 0)
 149                 SET_CONSOLE_3270;
 150 #endif
 151         set_preferred_console();
 152         return 1;
 153 }
 154
 155 __setup("conmode=", conmode_setup);
 156
 157 static void __init conmode_default(void)
 158 {
 159         char query_buffer[1024];
 160         char *ptr;
 161
 162         if (MACHINE_IS_VM) {
 163                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 164                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 165                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 166                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 167                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 168                 ptr = strstr(query_buffer, "CONMODE");
 169                 /*
 170                  * Set the conmode to 3215 so that the device recognition
 171                  * will set the cu_type of the console to 3215. If the
 172                  * conmode is 3270 and we don't set it back then both
 173                  * 3215 and the 3270 driver will try to access the console
 174                  * device (3215 as console and 3270 as normal tty).
 175                  */
 176                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 177                 if (ptr == NULL) {
 178 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 179                         SET_CONSOLE_SCLP;
 180 #endif
 181                         return;
 182                 }
 183                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 184 #if defined(CONFIG_TN3270_CONSOLE)
 185                         SET_CONSOLE_3270;
 186 #elif defined(CONFIG_TN3215_CONSOLE)
 187                         SET_CONSOLE_3215;
 188 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 189                         SET_CONSOLE_SCLP;
 190 #endif
 191                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 192 #if defined(CONFIG_TN3215_CONSOLE)
 193                         SET_CONSOLE_3215;
 194 #elif defined(CONFIG_TN3270_CONSOLE)
 195                         SET_CONSOLE_3270;
 196 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 197                         SET_CONSOLE_SCLP;
 198 #endif
 199                 }
 200         } else {
 201 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 202                 SET_CONSOLE_SCLP;
 203 #endif
 204         }
 205 }
 206
 207 #ifdef CONFIG_ZFCPDUMP
 208 static void __init setup_zfcpdump(unsigned int console_devno)
 209 {
 210         static char str[41];
 211
 212         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 213                 return;
 214         if (console_devno != -1)
 215                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 216                         ipl_info.data.fcp.dev_id.devno, console_devno);
 217         else
 218                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 219                         ipl_info.data.fcp.dev_id.devno);
 220         strcat(boot_command_line, str);
 221         console_loglevel = 2;
 222 }
 223 #else
 224 static inline void setup_zfcpdump(unsigned int console_devno) {}
 225 #endif /* CONFIG_ZFCPDUMP */
 226
 227  /*
 228  * Reboot, halt and power_off stubs. They just call _machine_restart,
 229  * _machine_halt or _machine_power_off.
 230  */
 231
 232 void machine_restart(char *command)
 233 {
 234         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 235                 /*
 236                  * Only unblank the console if we are called in enabled
 237                  * context or a bust_spinlocks cleared the way for us.
 238                  */
 239                 console_unblank();
 240         _machine_restart(command);
 241 }
 242
 243 void machine_halt(void)
 244 {
 245         if (!in_interrupt() || oops_in_progress)
 246                 /*
 247                  * Only unblank the console if we are called in enabled
 248                  * context or a bust_spinlocks cleared the way for us.
 249                  */
 250                 console_unblank();
 251         _machine_halt();
 252 }
 253
 254 void machine_power_off(void)
 255 {
 256         if (!in_interrupt() || oops_in_progress)
 257                 /*
 258                  * Only unblank the console if we are called in enabled
 259                  * context or a bust_spinlocks cleared the way for us.
 260                  */
 261                 console_unblank();
 262         _machine_power_off();
 263 }
 264
 265 /*
 266  * Dummy power off function.
 267  */
 268 void (*pm_power_off)(void) = machine_power_off;
 269
 270 static int __init early_parse_mem(char *p)
 271 {
 272         memory_end = memparse(p, &p);
 273         memory_end_set = 1;
 274         return 0;
 275 }
 276 early_param("mem", early_parse_mem);
 277
 278 unsigned int user_mode = HOME_SPACE_MODE;
 279 EXPORT_SYMBOL_GPL(user_mode);
 280
 281 static int set_amode_primary(void)
 282 {
 283         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
 284         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
 285 #ifdef CONFIG_COMPAT
 286         psw32_user_bits =
 287                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
 288 #endif
 289
 290         if (MACHINE_HAS_MVCOS) {
 291                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
 292                 return 1;
 293         } else {
 294                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
 295                 return 0;
 296         }
 297 }
 298
 299 /*
 300  * Switch kernel/user addressing modes?
 301  */
 302 static int __init early_parse_switch_amode(char *p)
 303 {
 304         user_mode = PRIMARY_SPACE_MODE;
 305         return 0;
 306 }
 307 early_param("switch_amode", early_parse_switch_amode);
 308
 309 static int __init early_parse_user_mode(char *p)
 310 {
 311         if (p && strcmp(p, "primary") == 0)
 312                 user_mode = PRIMARY_SPACE_MODE;
 313         else if (!p || strcmp(p, "home") == 0)
 314                 user_mode = HOME_SPACE_MODE;
 315         else
 316                 return 1;
 317         return 0;
 318 }
 319 early_param("user_mode", early_parse_user_mode);
 320
 321 static void setup_addressing_mode(void)
 322 {
 323         if (user_mode == PRIMARY_SPACE_MODE) {
 324                 if (set_amode_primary())
 325                         pr_info("Address spaces switched, "
 326                                 "mvcos available\n");
 327                 else
 328                         pr_info("Address spaces switched, "
 329                                 "mvcos not available\n");
 330         }
 331 }
 332
 333 static void __init
 334 setup_lowcore(void)
 335 {
 336         struct _lowcore *lc;
 337
 338         /*
 339          * Setup lowcore for boot cpu
 340          */
 341         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 342         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 343         lc->restart_psw.mask = psw_kernel_bits;
 344         lc->restart_psw.addr =
 345                 PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
 346         lc->external_new_psw.mask = psw_kernel_bits |
 347                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 348         lc->external_new_psw.addr =
 349                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 350         lc->svc_new_psw.mask = psw_kernel_bits |
 351                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 352         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 353         lc->program_new_psw.mask = psw_kernel_bits |
 354                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 355         lc->program_new_psw.addr =
 356                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 357         lc->mcck_new_psw.mask = psw_kernel_bits;
 358         lc->mcck_new_psw.addr =
 359                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 360         lc->io_new_psw.mask = psw_kernel_bits |
 361                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 362         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 363         lc->clock_comparator = -1ULL;
 364         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 365         lc->async_stack = (unsigned long)
 366                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 367         lc->panic_stack = (unsigned long)
 368                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 369         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 370         lc->thread_info = (unsigned long) &init_thread_union;
 371         lc->machine_flags = S390_lowcore.machine_flags;
 372         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 373         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 374                MAX_FACILITY_BIT/8);
 375 #ifndef CONFIG_64BIT
 376         if (MACHINE_HAS_IEEE) {
 377                 lc->extended_save_area_addr = (__u32)
 378                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 379                 /* enable extended save area */
 380                 __ctl_set_bit(14, 29);
 381         }
 382 #else
 383         lc->cmf_hpp = -1ULL;
 384         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 385 #endif
 386         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 387         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 388         lc->exit_timer = S390_lowcore.exit_timer;
 389         lc->user_timer = S390_lowcore.user_timer;
 390         lc->system_timer = S390_lowcore.system_timer;
 391         lc->steal_timer = S390_lowcore.steal_timer;
 392         lc->last_update_timer = S390_lowcore.last_update_timer;
 393         lc->last_update_clock = S390_lowcore.last_update_clock;
 394         lc->ftrace_func = S390_lowcore.ftrace_func;
 395         set_prefix((u32)(unsigned long) lc);
 396         lowcore_ptr[0] = lc;
 397 }
 398
 399 static struct resource code_resource = {
 400         .name  = "Kernel code",
 401         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 402 };
 403
 404 static struct resource data_resource = {
 405         .name = "Kernel data",
 406         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 407 };
 408
 409 static struct resource bss_resource = {
 410         .name = "Kernel bss",
 411         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 412 };
 413
 414 static struct resource __initdata *standard_resources[] = {
 415         &code_resource,
 416         &data_resource,
 417         &bss_resource,
 418 };
 419
 420 static void __init setup_resources(void)
 421 {
 422         struct resource *res, *std_res, *sub_res;
 423         int i, j;
 424
 425         code_resource.start = (unsigned long) &_text;
 426         code_resource.end = (unsigned long) &_etext - 1;
 427         data_resource.start = (unsigned long) &_etext;
 428         data_resource.end = (unsigned long) &_edata - 1;
 429         bss_resource.start = (unsigned long) &__bss_start;
 430         bss_resource.end = (unsigned long) &__bss_stop - 1;
 431
 432         for (i = 0; i < MEMORY_CHUNKS; i++) {
 433                 if (!memory_chunk[i].size)
 434                         continue;
 435                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
 436                     memory_chunk[i].type == CHUNK_CRASHK)
 437                         continue;
 438                 res = alloc_bootmem_low(sizeof(*res));
 439                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 440                 switch (memory_chunk[i].type) {
 441                 case CHUNK_READ_WRITE:
 442                 case CHUNK_CRASHK:
 443                         res->name = "System RAM";
 444                         break;
 445                 case CHUNK_READ_ONLY:
 446                         res->name = "System ROM";
 447                         res->flags |= IORESOURCE_READONLY;
 448                         break;
 449                 default:
 450                         res->name = "reserved";
 451                 }
 452                 res->start = memory_chunk[i].addr;
 453                 res->end = res->start + memory_chunk[i].size - 1;
 454                 request_resource(&iomem_resource, res);
 455
 456                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 457                         std_res = standard_resources[j];
 458                         if (std_res->start < res->start ||
 459                             std_res->start > res->end)
 460                                 continue;
 461                         if (std_res->end > res->end) {
 462                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 463                                 *sub_res = *std_res;
 464                                 sub_res->end = res->end;
 465                                 std_res->start = res->end + 1;
 466                                 request_resource(res, sub_res);
 467                         } else {
 468                                 request_resource(res, std_res);
 469                         }
 470                 }
 471         }
 472 }
 473
 474 unsigned long real_memory_size;
 475 EXPORT_SYMBOL_GPL(real_memory_size);
 476
 477 static void __init setup_memory_end(void)
 478 {
 479         unsigned long memory_size;
 480         unsigned long max_mem;
 481         int i;
 482
 483
 484 #ifdef CONFIG_ZFCPDUMP
 485         if (ipl_info.type == IPL_TYPE_FCP_DUMP) {
 486                 memory_end = ZFCPDUMP_HSA_SIZE;
 487                 memory_end_set = 1;
 488         }
 489 #endif
 490         memory_size = 0;
 491         memory_end &= PAGE_MASK;
 492
 493         max_mem = memory_end ? min(VMEM_MAX_PHYS, memory_end) : VMEM_MAX_PHYS;
 494         memory_end = min(max_mem, memory_end);
 495
 496         /*
 497          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 498          * extra checks that HOLES_IN_ZONE would require.
 499          */
 500         for (i = 0; i < MEMORY_CHUNKS; i++) {
 501                 unsigned long start, end;
 502                 struct mem_chunk *chunk;
 503                 unsigned long align;
 504
 505                 chunk = &memory_chunk[i];
 506                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 507                 start = (chunk->addr + align - 1) & ~(align - 1);
 508                 end = (chunk->addr + chunk->size) & ~(align - 1);
 509                 if (start >= end)
 510                         memset(chunk, 0, sizeof(*chunk));
 511                 else {
 512                         chunk->addr = start;
 513                         chunk->size = end - start;
 514                 }
 515         }
 516
 517         for (i = 0; i < MEMORY_CHUNKS; i++) {
 518                 struct mem_chunk *chunk = &memory_chunk[i];
 519
 520                 real_memory_size = max(real_memory_size,
 521                                        chunk->addr + chunk->size);
 522                 if (chunk->addr >= max_mem) {
 523                         memset(chunk, 0, sizeof(*chunk));
 524                         continue;
 525                 }
 526                 if (chunk->addr + chunk->size > max_mem)
 527                         chunk->size = max_mem - chunk->addr;
 528                 memory_size = max(memory_size, chunk->addr + chunk->size);
 529         }
 530         if (!memory_end)
 531                 memory_end = memory_size;
 532 }
 533
 534 void *restart_stack __attribute__((__section__(".data")));
 535
 536 /*
 537  * Setup new PSW and allocate stack for PSW restart interrupt
 538  */
 539 static void __init setup_restart_psw(void)
 540 {
 541         psw_t psw;
 542
 543         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 544         restart_stack += ASYNC_SIZE;
 545
 546         /*
 547          * Setup restart PSW for absolute zero lowcore. This is necesary
 548          * if PSW restart is done on an offline CPU that has lowcore zero
 549          */
 550         psw.mask = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_EA | PSW_MASK_BA;
 551         psw.addr = PSW_ADDR_AMODE | (unsigned long) psw_restart_int_handler;
 552         copy_to_absolute_zero(&S390_lowcore.restart_psw, &psw, sizeof(psw));
 553 }
 554
 555 static void __init setup_vmcoreinfo(void)
 556 {
 557 #ifdef CONFIG_KEXEC
 558         unsigned long ptr = paddr_vmcoreinfo_note();
 559
 560         copy_to_absolute_zero(&S390_lowcore.vmcore_info, &ptr, sizeof(ptr));
 561 #endif
 562 }
 563
 564 #ifdef CONFIG_CRASH_DUMP
 565
 566 /*
 567  * Find suitable location for crashkernel memory
 568  */
 569 static unsigned long __init find_crash_base(unsigned long crash_size,
 570                                             char **msg)
 571 {
 572         unsigned long crash_base;
 573         struct mem_chunk *chunk;
 574         int i;
 575
 576         if (memory_chunk[0].size < crash_size) {
 577                 *msg = "first memory chunk must be at least crashkernel size";
 578                 return 0;
 579         }
 580         if (is_kdump_kernel() && (crash_size == OLDMEM_SIZE))
 581                 return OLDMEM_BASE;
 582
 583         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
 584                 chunk = &memory_chunk[i];
 585                 if (chunk->size == 0)
 586                         continue;
 587                 if (chunk->type != CHUNK_READ_WRITE)
 588                         continue;
 589                 if (chunk->size < crash_size)
 590                         continue;
 591                 crash_base = (chunk->addr + chunk->size) - crash_size;
 592                 if (crash_base < crash_size)
 593                         continue;
 594                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
 595                         continue;
 596                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
 597                         continue;
 598                 return crash_base;
 599         }
 600         *msg = "no suitable area found";
 601         return 0;
 602 }
 603
 604 /*
 605  * Check if crash_base and crash_size is valid
 606  */
 607 static int __init verify_crash_base(unsigned long crash_base,
 608                                     unsigned long crash_size,
 609                                     char **msg)
 610 {
 611         struct mem_chunk *chunk;
 612         int i;
 613
 614         /*
 615          * Because we do the swap to zero, we must have at least 'crash_size'
 616          * bytes free space before crash_base
 617          */
 618         if (crash_size > crash_base) {
 619                 *msg = "crashkernel offset must be greater than size";
 620                 return -EINVAL;
 621         }
 622
 623         /* First memory chunk must be at least crash_size */
 624         if (memory_chunk[0].size < crash_size) {
 625                 *msg = "first memory chunk must be at least crashkernel size";
 626                 return -EINVAL;
 627         }
 628         /* Check if we fit into the respective memory chunk */
 629         for (i = 0; i < MEMORY_CHUNKS; i++) {
 630                 chunk = &memory_chunk[i];
 631                 if (chunk->size == 0)
 632                         continue;
 633                 if (crash_base < chunk->addr)
 634                         continue;
 635                 if (crash_base >= chunk->addr + chunk->size)
 636                         continue;
 637                 /* we have found the memory chunk */
 638                 if (crash_base + crash_size > chunk->addr + chunk->size) {
 639                         *msg = "selected memory chunk is too small for "
 640                                 "crashkernel memory";
 641                         return -EINVAL;
 642                 }
 643                 return 0;
 644         }
 645         *msg = "invalid memory range specified";
 646         return -EINVAL;
 647 }
 648
 649 /*
 650  * Reserve kdump memory by creating a memory hole in the mem_chunk array
 651  */
 652 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
 653                                          int type)
 654 {
 655
 656         create_mem_hole(memory_chunk, addr, size, type);
 657 }
 658
 659 /*
 660  * When kdump is enabled, we have to ensure that no memory from
 661  * the area [0 - crashkernel memory size] and
 662  * [crashk_res.start - crashk_res.end] is set offline.
 663  */
 664 static int kdump_mem_notifier(struct notifier_block *nb,
 665                               unsigned long action, void *data)
 666 {
 667         struct memory_notify *arg = data;
 668
 669         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 670                 return NOTIFY_BAD;
 671         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 672                 return NOTIFY_OK;
 673         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 674                 return NOTIFY_OK;
 675         return NOTIFY_BAD;
 676 }
 677
 678 static struct notifier_block kdump_mem_nb = {
 679         .notifier_call = kdump_mem_notifier,
 680 };
 681
 682 #endif
 683
 684 /*
 685  * Make sure that oldmem, where the dump is stored, is protected
 686  */
 687 static void reserve_oldmem(void)
 688 {
 689 #ifdef CONFIG_CRASH_DUMP
 690         if (!OLDMEM_BASE)
 691                 return;
 692
 693         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
 694         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
 695                               CHUNK_OLDMEM);
 696         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
 697                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
 698         else
 699                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
 700 #endif
 701 }
 702
 703 /*
 704  * Reserve memory for kdump kernel to be loaded with kexec
 705  */
 706 static void __init reserve_crashkernel(void)
 707 {
 708 #ifdef CONFIG_CRASH_DUMP
 709         unsigned long long crash_base, crash_size;
 710         char *msg;
 711         int rc;
 712
 713         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 714                                &crash_base);
 715         if (rc || crash_size == 0)
 716                 return;
 717         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 718         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 719         if (register_memory_notifier(&kdump_mem_nb))
 720                 return;
 721         if (!crash_base)
 722                 crash_base = find_crash_base(crash_size, &msg);
 723         if (!crash_base) {
 724                 pr_info("crashkernel reservation failed: %s\n", msg);
 725                 unregister_memory_notifier(&kdump_mem_nb);
 726                 return;
 727         }
 728         if (verify_crash_base(crash_base, crash_size, &msg)) {
 729                 pr_info("crashkernel reservation failed: %s\n", msg);
 730                 unregister_memory_notifier(&kdump_mem_nb);
 731                 return;
 732         }
 733         if (!OLDMEM_BASE && MACHINE_IS_VM)
 734                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 735         crashk_res.start = crash_base;
 736         crashk_res.end = crash_base + crash_size - 1;
 737         insert_resource(&iomem_resource, &crashk_res);
 738         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
 739         pr_info("Reserving %lluMB of memory at %lluMB "
 740                 "for crashkernel (System RAM: %luMB)\n",
 741                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
 742 #endif
 743 }
 744
 745 static void __init
 746 setup_memory(void)
 747 {
 748         unsigned long bootmap_size;
 749         unsigned long start_pfn, end_pfn;
 750         int i;
 751
 752         /*
 753          * partially used pages are not usable - thus
 754          * we are rounding upwards:
 755          */
 756         start_pfn = PFN_UP(__pa(&_end));
 757         end_pfn = max_pfn = PFN_DOWN(memory_end);
 758
 759 #ifdef CONFIG_BLK_DEV_INITRD
 760         /*
 761          * Move the initrd in case the bitmap of the bootmem allocater
 762          * would overwrite it.
 763          */
 764
 765         if (INITRD_START && INITRD_SIZE) {
 766                 unsigned long bmap_size;
 767                 unsigned long start;
 768
 769                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 770                 bmap_size = PFN_PHYS(bmap_size);
 771
 772                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 773                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 774
 775 #ifdef CONFIG_CRASH_DUMP
 776                         if (OLDMEM_BASE) {
 777                                 /* Move initrd behind kdump oldmem */
 778                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
 779                                     start < OLDMEM_BASE + OLDMEM_SIZE)
 780                                         start = OLDMEM_BASE + OLDMEM_SIZE;
 781                         }
 782 #endif
 783                         if (start + INITRD_SIZE > memory_end) {
 784                                 pr_err("initrd extends beyond end of "
 785                                        "memory (0x%08lx > 0x%08lx) "
 786                                        "disabling initrd\n",
 787                                        start + INITRD_SIZE, memory_end);
 788                                 INITRD_START = INITRD_SIZE = 0;
 789                         } else {
 790                                 pr_info("Moving initrd (0x%08lx -> "
 791                                         "0x%08lx, size: %ld)\n",
 792                                         INITRD_START, start, INITRD_SIZE);
 793                                 memmove((void *) start, (void *) INITRD_START,
 794                                         INITRD_SIZE);
 795                                 INITRD_START = start;
 796                         }
 797                 }
 798         }
 799 #endif
 800
 801         /*
 802          * Initialize the boot-time allocator
 803          */
 804         bootmap_size = init_bootmem(start_pfn, end_pfn);
 805
 806         /*
 807          * Register RAM areas with the bootmem allocator.
 808          */
 809
 810         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 811                 unsigned long start_chunk, end_chunk, pfn;
 812
 813                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
 814                     memory_chunk[i].type != CHUNK_CRASHK)
 815                         continue;
 816                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 817                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 818                 end_chunk = min(end_chunk, end_pfn);
 819                 if (start_chunk >= end_chunk)
 820                         continue;
 821                 add_active_range(0, start_chunk, end_chunk);
 822                 pfn = max(start_chunk, start_pfn);
 823                 for (; pfn < end_chunk; pfn++)
 824                         page_set_storage_key(PFN_PHYS(pfn),
 825                                              PAGE_DEFAULT_KEY, 0);
 826         }
 827
 828         psw_set_key(PAGE_DEFAULT_KEY);
 829
 830         free_bootmem_with_active_regions(0, max_pfn);
 831
 832         /*
 833          * Reserve memory used for lowcore/command line/kernel image.
 834          */
 835         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 836         reserve_bootmem((unsigned long)_stext,
 837                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 838                         BOOTMEM_DEFAULT);
 839         /*
 840          * Reserve the bootmem bitmap itself as well. We do this in two
 841          * steps (first step was init_bootmem()) because this catches
 842          * the (very unlikely) case of us accidentally initializing the
 843          * bootmem allocator with an invalid RAM area.
 844          */
 845         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 846                         BOOTMEM_DEFAULT);
 847
 848 #ifdef CONFIG_CRASH_DUMP
 849         if (crashk_res.start)
 850                 reserve_bootmem(crashk_res.start,
 851                                 crashk_res.end - crashk_res.start + 1,
 852                                 BOOTMEM_DEFAULT);
 853         if (is_kdump_kernel())
 854                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
 855                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
 856 #endif
 857 #ifdef CONFIG_BLK_DEV_INITRD
 858         if (INITRD_START && INITRD_SIZE) {
 859                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 860                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 861                                         BOOTMEM_DEFAULT);
 862                         initrd_start = INITRD_START;
 863                         initrd_end = initrd_start + INITRD_SIZE;
 864                 } else {
 865                         pr_err("initrd extends beyond end of "
 866                                "memory (0x%08lx > 0x%08lx) "
 867                                "disabling initrd\n",
 868                                initrd_start + INITRD_SIZE, memory_end);
 869                         initrd_start = initrd_end = 0;
 870                 }
 871         }
 872 #endif
 873 }
 874
 875 /*
 876  * Setup hardware capabilities.
 877  */
 878 static void __init setup_hwcaps(void)
 879 {
 880         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 881         struct cpuid cpu_id;
 882         int i;
 883
 884         /*
 885          * The store facility list bits numbers as found in the principles
 886          * of operation are numbered with bit 1UL<<31 as number 0 to
 887          * bit 1UL<<0 as number 31.
 888          *   Bit 0: instructions named N3, "backported" to esa-mode
 889          *   Bit 2: z/Architecture mode is active
 890          *   Bit 7: the store-facility-list-extended facility is installed
 891          *   Bit 17: the message-security assist is installed
 892          *   Bit 19: the long-displacement facility is installed
 893          *   Bit 21: the extended-immediate facility is installed
 894          *   Bit 22: extended-translation facility 3 is installed
 895          *   Bit 30: extended-translation facility 3 enhancement facility
 896          * These get translated to:
 897          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 898          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 899          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 900          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 901          */
 902         for (i = 0; i < 6; i++)
 903                 if (test_facility(stfl_bits[i]))
 904                         elf_hwcap |= 1UL << i;
 905
 906         if (test_facility(22) && test_facility(30))
 907                 elf_hwcap |= HWCAP_S390_ETF3EH;
 908
 909         /*
 910          * Check for additional facilities with store-facility-list-extended.
 911          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 912          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 913          * as stored by stfl, bits 32-xxx contain additional facilities.
 914          * How many facility words are stored depends on the number of
 915          * doublewords passed to the instruction. The additional facilities
 916          * are:
 917          *   Bit 42: decimal floating point facility is installed
 918          *   Bit 44: perform floating point operation facility is installed
 919          * translated to:
 920          *   HWCAP_S390_DFP bit 6 (42 && 44).
 921          */
 922         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 923                 elf_hwcap |= HWCAP_S390_DFP;
 924
 925         /*
 926          * Huge page support HWCAP_S390_HPAGE is bit 7.
 927          */
 928         if (MACHINE_HAS_HPAGE)
 929                 elf_hwcap |= HWCAP_S390_HPAGE;
 930
 931         /*
 932          * 64-bit register support for 31-bit processes
 933          * HWCAP_S390_HIGH_GPRS is bit 9.
 934          */
 935         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 936
 937         get_cpu_id(&cpu_id);
 938         switch (cpu_id.machine) {
 939         case 0x9672:
 940 #if !defined(CONFIG_64BIT)
 941         default:        /* Use "g5" as default for 31 bit kernels. */
 942 #endif
 943                 strcpy(elf_platform, "g5");
 944                 break;
 945         case 0x2064:
 946         case 0x2066:
 947 #if defined(CONFIG_64BIT)
 948         default:        /* Use "z900" as default for 64 bit kernels. */
 949 #endif
 950                 strcpy(elf_platform, "z900");
 951                 break;
 952         case 0x2084:
 953         case 0x2086:
 954                 strcpy(elf_platform, "z990");
 955                 break;
 956         case 0x2094:
 957         case 0x2096:
 958                 strcpy(elf_platform, "z9-109");
 959                 break;
 960         case 0x2097:
 961         case 0x2098:
 962                 strcpy(elf_platform, "z10");
 963                 break;
 964         case 0x2817:
 965         case 0x2818:
 966                 strcpy(elf_platform, "z196");
 967                 break;
 968         }
 969 }
 970
 971 /*
 972  * Setup function called from init/main.c just after the banner
 973  * was printed.
 974  */
 975
 976 void __init
 977 setup_arch(char **cmdline_p)
 978 {
 979         /*
 980          * print what head.S has found out about the machine
 981          */
 982 #ifndef CONFIG_64BIT
 983         if (MACHINE_IS_VM)
 984                 pr_info("Linux is running as a z/VM "
 985                         "guest operating system in 31-bit mode\n");
 986         else if (MACHINE_IS_LPAR)
 987                 pr_info("Linux is running natively in 31-bit mode\n");
 988         if (MACHINE_HAS_IEEE)
 989                 pr_info("The hardware system has IEEE compatible "
 990                         "floating point units\n");
 991         else
 992                 pr_info("The hardware system has no IEEE compatible "
 993                         "floating point units\n");
 994 #else /* CONFIG_64BIT */
 995         if (MACHINE_IS_VM)
 996                 pr_info("Linux is running as a z/VM "
 997                         "guest operating system in 64-bit mode\n");
 998         else if (MACHINE_IS_KVM)
 999                 pr_info("Linux is running under KVM in 64-bit mode\n");
1000         else if (MACHINE_IS_LPAR)
1001                 pr_info("Linux is running natively in 64-bit mode\n");
1002 #endif /* CONFIG_64BIT */
1003
1004         /* Have one command line that is parsed and saved in /proc/cmdline */
1005         /* boot_command_line has been already set up in early.c */
1006         *cmdline_p = boot_command_line;
1007
1008         ROOT_DEV = Root_RAM0;
1009
1010         init_mm.start_code = PAGE_OFFSET;
1011         init_mm.end_code = (unsigned long) &_etext;
1012         init_mm.end_data = (unsigned long) &_edata;
1013         init_mm.brk = (unsigned long) &_end;
1014
1015         if (MACHINE_HAS_MVCOS)
1016                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1017         else
1018                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1019
1020         parse_early_param();
1021
1022         setup_ipl();
1023         setup_memory_end();
1024         setup_addressing_mode();
1025         reserve_oldmem();
1026         reserve_crashkernel();
1027         setup_memory();
1028         setup_resources();
1029         setup_vmcoreinfo();
1030         setup_restart_psw();
1031         setup_lowcore();
1032
1033         cpu_init();
1034         s390_init_cpu_topology();
1035
1036         /*
1037          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1038          */
1039         setup_hwcaps();
1040
1041         /*
1042          * Create kernel page tables and switch to virtual addressing.
1043          */
1044         paging_init();
1045
1046         /* Setup default console */
1047         conmode_default();
1048         set_preferred_console();
1049
1050         /* Setup zfcpdump support */
1051         setup_zfcpdump(console_devno);
1052 }