arch/s390/kernel/setup.c

   1 /*
   2  *  arch/s390/kernel/setup.c
   3  *
   4  *  S390 version
   5  *    Copyright (C) IBM Corp. 1999,2012
   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/memblock.h>
  25 #include <linux/mm.h>
  26 #include <linux/stddef.h>
  27 #include <linux/unistd.h>
  28 #include <linux/ptrace.h>
  29 #include <linux/user.h>
  30 #include <linux/tty.h>
  31 #include <linux/ioport.h>
  32 #include <linux/delay.h>
  33 #include <linux/init.h>
  34 #include <linux/initrd.h>
  35 #include <linux/bootmem.h>
  36 #include <linux/root_dev.h>
  37 #include <linux/console.h>
  38 #include <linux/kernel_stat.h>
  39 #include <linux/device.h>
  40 #include <linux/notifier.h>
  41 #include <linux/pfn.h>
  42 #include <linux/ctype.h>
  43 #include <linux/reboot.h>
  44 #include <linux/topology.h>
  45 #include <linux/ftrace.h>
  46 #include <linux/kexec.h>
  47 #include <linux/crash_dump.h>
  48 #include <linux/memory.h>
  49 #include <linux/compat.h>
  50
  51 #include <asm/ipl.h>
  52 #include <asm/uaccess.h>
  53 #include <asm/facility.h>
  54 #include <asm/smp.h>
  55 #include <asm/mmu_context.h>
  56 #include <asm/cpcmd.h>
  57 #include <asm/lowcore.h>
  58 #include <asm/irq.h>
  59 #include <asm/page.h>
  60 #include <asm/ptrace.h>
  61 #include <asm/sections.h>
  62 #include <asm/ebcdic.h>
  63 #include <asm/kvm_virtio.h>
  64 #include <asm/diag.h>
  65 #include <asm/os_info.h>
  66 #include "entry.h"
  67
  68 long psw_kernel_bits    = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
  69                           PSW_MASK_EA | PSW_MASK_BA;
  70 long psw_user_bits      = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
  71                           PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
  72                           PSW_MASK_PSTATE | PSW_ASC_HOME;
  73
  74 /*
  75  * User copy operations.
  76  */
  77 struct uaccess_ops uaccess;
  78 EXPORT_SYMBOL(uaccess);
  79
  80 /*
  81  * Machine setup..
  82  */
  83 unsigned int console_mode = 0;
  84 EXPORT_SYMBOL(console_mode);
  85
  86 unsigned int console_devno = -1;
  87 EXPORT_SYMBOL(console_devno);
  88
  89 unsigned int console_irq = -1;
  90 EXPORT_SYMBOL(console_irq);
  91
  92 unsigned long elf_hwcap = 0;
  93 char elf_platform[ELF_PLATFORM_SIZE];
  94
  95 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
  96
  97 int __initdata memory_end_set;
  98 unsigned long __initdata memory_end;
  99
 100 unsigned long VMALLOC_START;
 101 EXPORT_SYMBOL(VMALLOC_START);
 102
 103 unsigned long VMALLOC_END;
 104 EXPORT_SYMBOL(VMALLOC_END);
 105
 106 struct page *vmemmap;
 107 EXPORT_SYMBOL(vmemmap);
 108
 109 /* An array with a pointer to the lowcore of every CPU. */
 110 struct _lowcore *lowcore_ptr[NR_CPUS];
 111 EXPORT_SYMBOL(lowcore_ptr);
 112
 113 /*
 114  * This is set up by the setup-routine at boot-time
 115  * for S390 need to find out, what we have to setup
 116  * using address 0x10400 ...
 117  */
 118
 119 #include <asm/setup.h>
 120
 121 /*
 122  * condev= and conmode= setup parameter.
 123  */
 124
 125 static int __init condev_setup(char *str)
 126 {
 127         int vdev;
 128
 129         vdev = simple_strtoul(str, &str, 0);
 130         if (vdev >= 0 && vdev < 65536) {
 131                 console_devno = vdev;
 132                 console_irq = -1;
 133         }
 134         return 1;
 135 }
 136
 137 __setup("condev=", condev_setup);
 138
 139 static void __init set_preferred_console(void)
 140 {
 141         if (MACHINE_IS_KVM)
 142                 add_preferred_console("hvc", 0, NULL);
 143         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 144                 add_preferred_console("ttyS", 0, NULL);
 145         else if (CONSOLE_IS_3270)
 146                 add_preferred_console("tty3270", 0, NULL);
 147 }
 148
 149 static int __init conmode_setup(char *str)
 150 {
 151 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 152         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 153                 SET_CONSOLE_SCLP;
 154 #endif
 155 #if defined(CONFIG_TN3215_CONSOLE)
 156         if (strncmp(str, "3215", 5) == 0)
 157                 SET_CONSOLE_3215;
 158 #endif
 159 #if defined(CONFIG_TN3270_CONSOLE)
 160         if (strncmp(str, "3270", 5) == 0)
 161                 SET_CONSOLE_3270;
 162 #endif
 163         set_preferred_console();
 164         return 1;
 165 }
 166
 167 __setup("conmode=", conmode_setup);
 168
 169 static void __init conmode_default(void)
 170 {
 171         char query_buffer[1024];
 172         char *ptr;
 173
 174         if (MACHINE_IS_VM) {
 175                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 176                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 177                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 178                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 179                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 180                 ptr = strstr(query_buffer, "CONMODE");
 181                 /*
 182                  * Set the conmode to 3215 so that the device recognition
 183                  * will set the cu_type of the console to 3215. If the
 184                  * conmode is 3270 and we don't set it back then both
 185                  * 3215 and the 3270 driver will try to access the console
 186                  * device (3215 as console and 3270 as normal tty).
 187                  */
 188                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 189                 if (ptr == NULL) {
 190 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 191                         SET_CONSOLE_SCLP;
 192 #endif
 193                         return;
 194                 }
 195                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 196 #if defined(CONFIG_TN3270_CONSOLE)
 197                         SET_CONSOLE_3270;
 198 #elif defined(CONFIG_TN3215_CONSOLE)
 199                         SET_CONSOLE_3215;
 200 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 201                         SET_CONSOLE_SCLP;
 202 #endif
 203                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 204 #if defined(CONFIG_TN3215_CONSOLE)
 205                         SET_CONSOLE_3215;
 206 #elif defined(CONFIG_TN3270_CONSOLE)
 207                         SET_CONSOLE_3270;
 208 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 209                         SET_CONSOLE_SCLP;
 210 #endif
 211                 }
 212         } else {
 213 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 214                 SET_CONSOLE_SCLP;
 215 #endif
 216         }
 217 }
 218
 219 #ifdef CONFIG_ZFCPDUMP
 220 static void __init setup_zfcpdump(unsigned int console_devno)
 221 {
 222         static char str[41];
 223
 224         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 225                 return;
 226         if (OLDMEM_BASE)
 227                 return;
 228         if (console_devno != -1)
 229                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 230                         ipl_info.data.fcp.dev_id.devno, console_devno);
 231         else
 232                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 233                         ipl_info.data.fcp.dev_id.devno);
 234         strcat(boot_command_line, str);
 235         console_loglevel = 2;
 236 }
 237 #else
 238 static inline void setup_zfcpdump(unsigned int console_devno) {}
 239 #endif /* CONFIG_ZFCPDUMP */
 240
 241  /*
 242  * Reboot, halt and power_off stubs. They just call _machine_restart,
 243  * _machine_halt or _machine_power_off.
 244  */
 245
 246 void machine_restart(char *command)
 247 {
 248         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 249                 /*
 250                  * Only unblank the console if we are called in enabled
 251                  * context or a bust_spinlocks cleared the way for us.
 252                  */
 253                 console_unblank();
 254         _machine_restart(command);
 255 }
 256
 257 void machine_halt(void)
 258 {
 259         if (!in_interrupt() || oops_in_progress)
 260                 /*
 261                  * Only unblank the console if we are called in enabled
 262                  * context or a bust_spinlocks cleared the way for us.
 263                  */
 264                 console_unblank();
 265         _machine_halt();
 266 }
 267
 268 void machine_power_off(void)
 269 {
 270         if (!in_interrupt() || oops_in_progress)
 271                 /*
 272                  * Only unblank the console if we are called in enabled
 273                  * context or a bust_spinlocks cleared the way for us.
 274                  */
 275                 console_unblank();
 276         _machine_power_off();
 277 }
 278
 279 /*
 280  * Dummy power off function.
 281  */
 282 void (*pm_power_off)(void) = machine_power_off;
 283
 284 static int __init early_parse_mem(char *p)
 285 {
 286         memory_end = memparse(p, &p);
 287         memory_end_set = 1;
 288         return 0;
 289 }
 290 early_param("mem", early_parse_mem);
 291
 292 static int __init parse_vmalloc(char *arg)
 293 {
 294         if (!arg)
 295                 return -EINVAL;
 296         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 297         return 0;
 298 }
 299 early_param("vmalloc", parse_vmalloc);
 300
 301 unsigned int user_mode = HOME_SPACE_MODE;
 302 EXPORT_SYMBOL_GPL(user_mode);
 303
 304 static int set_amode_primary(void)
 305 {
 306         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
 307         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
 308 #ifdef CONFIG_COMPAT
 309         psw32_user_bits =
 310                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
 311 #endif
 312
 313         if (MACHINE_HAS_MVCOS) {
 314                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
 315                 return 1;
 316         } else {
 317                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
 318                 return 0;
 319         }
 320 }
 321
 322 /*
 323  * Switch kernel/user addressing modes?
 324  */
 325 static int __init early_parse_switch_amode(char *p)
 326 {
 327         user_mode = PRIMARY_SPACE_MODE;
 328         return 0;
 329 }
 330 early_param("switch_amode", early_parse_switch_amode);
 331
 332 static int __init early_parse_user_mode(char *p)
 333 {
 334         if (p && strcmp(p, "primary") == 0)
 335                 user_mode = PRIMARY_SPACE_MODE;
 336         else if (!p || strcmp(p, "home") == 0)
 337                 user_mode = HOME_SPACE_MODE;
 338         else
 339                 return 1;
 340         return 0;
 341 }
 342 early_param("user_mode", early_parse_user_mode);
 343
 344 static void setup_addressing_mode(void)
 345 {
 346         if (user_mode == PRIMARY_SPACE_MODE) {
 347                 if (set_amode_primary())
 348                         pr_info("Address spaces switched, "
 349                                 "mvcos available\n");
 350                 else
 351                         pr_info("Address spaces switched, "
 352                                 "mvcos not available\n");
 353         }
 354 }
 355
 356 void *restart_stack __attribute__((__section__(".data")));
 357
 358 static void __init setup_lowcore(void)
 359 {
 360         struct _lowcore *lc;
 361
 362         /*
 363          * Setup lowcore for boot cpu
 364          */
 365         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 366         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 367         lc->restart_psw.mask = psw_kernel_bits;
 368         lc->restart_psw.addr =
 369                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 370         lc->external_new_psw.mask = psw_kernel_bits |
 371                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 372         lc->external_new_psw.addr =
 373                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 374         lc->svc_new_psw.mask = psw_kernel_bits |
 375                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 376         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 377         lc->program_new_psw.mask = psw_kernel_bits |
 378                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 379         lc->program_new_psw.addr =
 380                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 381         lc->mcck_new_psw.mask = psw_kernel_bits;
 382         lc->mcck_new_psw.addr =
 383                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 384         lc->io_new_psw.mask = psw_kernel_bits |
 385                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 386         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 387         lc->clock_comparator = -1ULL;
 388         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 389         lc->async_stack = (unsigned long)
 390                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 391         lc->panic_stack = (unsigned long)
 392                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 393         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 394         lc->thread_info = (unsigned long) &init_thread_union;
 395         lc->machine_flags = S390_lowcore.machine_flags;
 396         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 397         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 398                MAX_FACILITY_BIT/8);
 399 #ifndef CONFIG_64BIT
 400         if (MACHINE_HAS_IEEE) {
 401                 lc->extended_save_area_addr = (__u32)
 402                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 403                 /* enable extended save area */
 404                 __ctl_set_bit(14, 29);
 405         }
 406 #else
 407         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 408 #endif
 409         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 410         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 411         lc->exit_timer = S390_lowcore.exit_timer;
 412         lc->user_timer = S390_lowcore.user_timer;
 413         lc->system_timer = S390_lowcore.system_timer;
 414         lc->steal_timer = S390_lowcore.steal_timer;
 415         lc->last_update_timer = S390_lowcore.last_update_timer;
 416         lc->last_update_clock = S390_lowcore.last_update_clock;
 417         lc->ftrace_func = S390_lowcore.ftrace_func;
 418
 419         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 420         restart_stack += ASYNC_SIZE;
 421
 422         /*
 423          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 424          * restart data to the absolute zero lowcore. This is necesary if
 425          * PSW restart is done on an offline CPU that has lowcore zero.
 426          */
 427         lc->restart_stack = (unsigned long) restart_stack;
 428         lc->restart_fn = (unsigned long) do_restart;
 429         lc->restart_data = 0;
 430         lc->restart_source = -1UL;
 431
 432         /* Setup absolute zero lowcore */
 433         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 434         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 435         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 436         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 437         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 438
 439         set_prefix((u32)(unsigned long) lc);
 440         lowcore_ptr[0] = lc;
 441 }
 442
 443 static struct resource code_resource = {
 444         .name  = "Kernel code",
 445         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 446 };
 447
 448 static struct resource data_resource = {
 449         .name = "Kernel data",
 450         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 451 };
 452
 453 static struct resource bss_resource = {
 454         .name = "Kernel bss",
 455         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 456 };
 457
 458 static struct resource __initdata *standard_resources[] = {
 459         &code_resource,
 460         &data_resource,
 461         &bss_resource,
 462 };
 463
 464 static void __init setup_resources(void)
 465 {
 466         struct resource *res, *std_res, *sub_res;
 467         int i, j;
 468
 469         code_resource.start = (unsigned long) &_text;
 470         code_resource.end = (unsigned long) &_etext - 1;
 471         data_resource.start = (unsigned long) &_etext;
 472         data_resource.end = (unsigned long) &_edata - 1;
 473         bss_resource.start = (unsigned long) &__bss_start;
 474         bss_resource.end = (unsigned long) &__bss_stop - 1;
 475
 476         for (i = 0; i < MEMORY_CHUNKS; i++) {
 477                 if (!memory_chunk[i].size)
 478                         continue;
 479                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
 480                     memory_chunk[i].type == CHUNK_CRASHK)
 481                         continue;
 482                 res = alloc_bootmem_low(sizeof(*res));
 483                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 484                 switch (memory_chunk[i].type) {
 485                 case CHUNK_READ_WRITE:
 486                 case CHUNK_CRASHK:
 487                         res->name = "System RAM";
 488                         break;
 489                 case CHUNK_READ_ONLY:
 490                         res->name = "System ROM";
 491                         res->flags |= IORESOURCE_READONLY;
 492                         break;
 493                 default:
 494                         res->name = "reserved";
 495                 }
 496                 res->start = memory_chunk[i].addr;
 497                 res->end = res->start + memory_chunk[i].size - 1;
 498                 request_resource(&iomem_resource, res);
 499
 500                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 501                         std_res = standard_resources[j];
 502                         if (std_res->start < res->start ||
 503                             std_res->start > res->end)
 504                                 continue;
 505                         if (std_res->end > res->end) {
 506                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 507                                 *sub_res = *std_res;
 508                                 sub_res->end = res->end;
 509                                 std_res->start = res->end + 1;
 510                                 request_resource(res, sub_res);
 511                         } else {
 512                                 request_resource(res, std_res);
 513                         }
 514                 }
 515         }
 516 }
 517
 518 unsigned long real_memory_size;
 519 EXPORT_SYMBOL_GPL(real_memory_size);
 520
 521 static void __init setup_memory_end(void)
 522 {
 523         unsigned long vmax, vmalloc_size, tmp;
 524         int i;
 525
 526
 527 #ifdef CONFIG_ZFCPDUMP
 528         if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
 529                 memory_end = ZFCPDUMP_HSA_SIZE;
 530                 memory_end_set = 1;
 531         }
 532 #endif
 533         real_memory_size = 0;
 534         memory_end &= PAGE_MASK;
 535
 536         /*
 537          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 538          * extra checks that HOLES_IN_ZONE would require.
 539          */
 540         for (i = 0; i < MEMORY_CHUNKS; i++) {
 541                 unsigned long start, end;
 542                 struct mem_chunk *chunk;
 543                 unsigned long align;
 544
 545                 chunk = &memory_chunk[i];
 546                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 547                 start = (chunk->addr + align - 1) & ~(align - 1);
 548                 end = (chunk->addr + chunk->size) & ~(align - 1);
 549                 if (start >= end)
 550                         memset(chunk, 0, sizeof(*chunk));
 551                 else {
 552                         chunk->addr = start;
 553                         chunk->size = end - start;
 554                 }
 555                 real_memory_size = max(real_memory_size,
 556                                        chunk->addr + chunk->size);
 557         }
 558
 559         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 560 #ifdef CONFIG_64BIT
 561         vmalloc_size = VMALLOC_END ?: 128UL << 30;
 562         tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
 563         tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
 564         if (tmp <= (1UL << 42))
 565                 vmax = 1UL << 42;       /* 3-level kernel page table */
 566         else
 567                 vmax = 1UL << 53;       /* 4-level kernel page table */
 568 #else
 569         vmalloc_size = VMALLOC_END ?: 96UL << 20;
 570         vmax = 1UL << 31;               /* 2-level kernel page table */
 571 #endif
 572         /* vmalloc area is at the end of the kernel address space. */
 573         VMALLOC_END = vmax;
 574         VMALLOC_START = vmax - vmalloc_size;
 575
 576         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 577         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 578         tmp = VMALLOC_START - tmp * sizeof(struct page);
 579         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 580         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 581         vmemmap = (struct page *) tmp;
 582
 583         /* Take care that memory_end is set and <= vmemmap */
 584         memory_end = min(memory_end ?: real_memory_size, tmp);
 585
 586         /* Fixup memory chunk array to fit into 0..memory_end */
 587         for (i = 0; i < MEMORY_CHUNKS; i++) {
 588                 struct mem_chunk *chunk = &memory_chunk[i];
 589
 590                 if (chunk->addr >= memory_end) {
 591                         memset(chunk, 0, sizeof(*chunk));
 592                         continue;
 593                 }
 594                 if (chunk->addr + chunk->size > memory_end)
 595                         chunk->size = memory_end - chunk->addr;
 596         }
 597 }
 598
 599 static void __init setup_vmcoreinfo(void)
 600 {
 601 #ifdef CONFIG_KEXEC
 602         mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
 603 #endif
 604 }
 605
 606 #ifdef CONFIG_CRASH_DUMP
 607
 608 /*
 609  * Find suitable location for crashkernel memory
 610  */
 611 static unsigned long __init find_crash_base(unsigned long crash_size,
 612                                             char **msg)
 613 {
 614         unsigned long crash_base;
 615         struct mem_chunk *chunk;
 616         int i;
 617
 618         if (memory_chunk[0].size < crash_size) {
 619                 *msg = "first memory chunk must be at least crashkernel size";
 620                 return 0;
 621         }
 622         if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
 623                 return OLDMEM_BASE;
 624
 625         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
 626                 chunk = &memory_chunk[i];
 627                 if (chunk->size == 0)
 628                         continue;
 629                 if (chunk->type != CHUNK_READ_WRITE)
 630                         continue;
 631                 if (chunk->size < crash_size)
 632                         continue;
 633                 crash_base = (chunk->addr + chunk->size) - crash_size;
 634                 if (crash_base < crash_size)
 635                         continue;
 636                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
 637                         continue;
 638                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
 639                         continue;
 640                 return crash_base;
 641         }
 642         *msg = "no suitable area found";
 643         return 0;
 644 }
 645
 646 /*
 647  * Check if crash_base and crash_size is valid
 648  */
 649 static int __init verify_crash_base(unsigned long crash_base,
 650                                     unsigned long crash_size,
 651                                     char **msg)
 652 {
 653         struct mem_chunk *chunk;
 654         int i;
 655
 656         /*
 657          * Because we do the swap to zero, we must have at least 'crash_size'
 658          * bytes free space before crash_base
 659          */
 660         if (crash_size > crash_base) {
 661                 *msg = "crashkernel offset must be greater than size";
 662                 return -EINVAL;
 663         }
 664
 665         /* First memory chunk must be at least crash_size */
 666         if (memory_chunk[0].size < crash_size) {
 667                 *msg = "first memory chunk must be at least crashkernel size";
 668                 return -EINVAL;
 669         }
 670         /* Check if we fit into the respective memory chunk */
 671         for (i = 0; i < MEMORY_CHUNKS; i++) {
 672                 chunk = &memory_chunk[i];
 673                 if (chunk->size == 0)
 674                         continue;
 675                 if (crash_base < chunk->addr)
 676                         continue;
 677                 if (crash_base >= chunk->addr + chunk->size)
 678                         continue;
 679                 /* we have found the memory chunk */
 680                 if (crash_base + crash_size > chunk->addr + chunk->size) {
 681                         *msg = "selected memory chunk is too small for "
 682                                 "crashkernel memory";
 683                         return -EINVAL;
 684                 }
 685                 return 0;
 686         }
 687         *msg = "invalid memory range specified";
 688         return -EINVAL;
 689 }
 690
 691 /*
 692  * Reserve kdump memory by creating a memory hole in the mem_chunk array
 693  */
 694 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
 695                                          int type)
 696 {
 697         create_mem_hole(memory_chunk, addr, size, type);
 698 }
 699
 700 /*
 701  * When kdump is enabled, we have to ensure that no memory from
 702  * the area [0 - crashkernel memory size] and
 703  * [crashk_res.start - crashk_res.end] is set offline.
 704  */
 705 static int kdump_mem_notifier(struct notifier_block *nb,
 706                               unsigned long action, void *data)
 707 {
 708         struct memory_notify *arg = data;
 709
 710         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 711                 return NOTIFY_BAD;
 712         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 713                 return NOTIFY_OK;
 714         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 715                 return NOTIFY_OK;
 716         return NOTIFY_BAD;
 717 }
 718
 719 static struct notifier_block kdump_mem_nb = {
 720         .notifier_call = kdump_mem_notifier,
 721 };
 722
 723 #endif
 724
 725 /*
 726  * Make sure that oldmem, where the dump is stored, is protected
 727  */
 728 static void reserve_oldmem(void)
 729 {
 730 #ifdef CONFIG_CRASH_DUMP
 731         if (!OLDMEM_BASE)
 732                 return;
 733
 734         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
 735         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
 736                               CHUNK_OLDMEM);
 737         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
 738                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
 739         else
 740                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
 741 #endif
 742 }
 743
 744 /*
 745  * Reserve memory for kdump kernel to be loaded with kexec
 746  */
 747 static void __init reserve_crashkernel(void)
 748 {
 749 #ifdef CONFIG_CRASH_DUMP
 750         unsigned long long crash_base, crash_size;
 751         char *msg = NULL;
 752         int rc;
 753
 754         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 755                                &crash_base);
 756         if (rc || crash_size == 0)
 757                 return;
 758         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 759         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 760         if (register_memory_notifier(&kdump_mem_nb))
 761                 return;
 762         if (!crash_base)
 763                 crash_base = find_crash_base(crash_size, &msg);
 764         if (!crash_base) {
 765                 pr_info("crashkernel reservation failed: %s\n", msg);
 766                 unregister_memory_notifier(&kdump_mem_nb);
 767                 return;
 768         }
 769         if (verify_crash_base(crash_base, crash_size, &msg)) {
 770                 pr_info("crashkernel reservation failed: %s\n", msg);
 771                 unregister_memory_notifier(&kdump_mem_nb);
 772                 return;
 773         }
 774         if (!OLDMEM_BASE && MACHINE_IS_VM)
 775                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 776         crashk_res.start = crash_base;
 777         crashk_res.end = crash_base + crash_size - 1;
 778         insert_resource(&iomem_resource, &crashk_res);
 779         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
 780         pr_info("Reserving %lluMB of memory at %lluMB "
 781                 "for crashkernel (System RAM: %luMB)\n",
 782                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
 783         os_info_crashkernel_add(crash_base, crash_size);
 784 #endif
 785 }
 786
 787 static void __init setup_memory(void)
 788 {
 789         unsigned long bootmap_size;
 790         unsigned long start_pfn, end_pfn;
 791         int i;
 792
 793         /*
 794          * partially used pages are not usable - thus
 795          * we are rounding upwards:
 796          */
 797         start_pfn = PFN_UP(__pa(&_end));
 798         end_pfn = max_pfn = PFN_DOWN(memory_end);
 799
 800 #ifdef CONFIG_BLK_DEV_INITRD
 801         /*
 802          * Move the initrd in case the bitmap of the bootmem allocater
 803          * would overwrite it.
 804          */
 805
 806         if (INITRD_START && INITRD_SIZE) {
 807                 unsigned long bmap_size;
 808                 unsigned long start;
 809
 810                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 811                 bmap_size = PFN_PHYS(bmap_size);
 812
 813                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 814                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 815
 816 #ifdef CONFIG_CRASH_DUMP
 817                         if (OLDMEM_BASE) {
 818                                 /* Move initrd behind kdump oldmem */
 819                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
 820                                     start < OLDMEM_BASE + OLDMEM_SIZE)
 821                                         start = OLDMEM_BASE + OLDMEM_SIZE;
 822                         }
 823 #endif
 824                         if (start + INITRD_SIZE > memory_end) {
 825                                 pr_err("initrd extends beyond end of "
 826                                        "memory (0x%08lx > 0x%08lx) "
 827                                        "disabling initrd\n",
 828                                        start + INITRD_SIZE, memory_end);
 829                                 INITRD_START = INITRD_SIZE = 0;
 830                         } else {
 831                                 pr_info("Moving initrd (0x%08lx -> "
 832                                         "0x%08lx, size: %ld)\n",
 833                                         INITRD_START, start, INITRD_SIZE);
 834                                 memmove((void *) start, (void *) INITRD_START,
 835                                         INITRD_SIZE);
 836                                 INITRD_START = start;
 837                         }
 838                 }
 839         }
 840 #endif
 841
 842         /*
 843          * Initialize the boot-time allocator
 844          */
 845         bootmap_size = init_bootmem(start_pfn, end_pfn);
 846
 847         /*
 848          * Register RAM areas with the bootmem allocator.
 849          */
 850
 851         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 852                 unsigned long start_chunk, end_chunk, pfn;
 853
 854                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
 855                     memory_chunk[i].type != CHUNK_CRASHK)
 856                         continue;
 857                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 858                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 859                 end_chunk = min(end_chunk, end_pfn);
 860                 if (start_chunk >= end_chunk)
 861                         continue;
 862                 memblock_add_node(PFN_PHYS(start_chunk),
 863                                   PFN_PHYS(end_chunk - start_chunk), 0);
 864                 pfn = max(start_chunk, start_pfn);
 865                 for (; pfn < end_chunk; pfn++)
 866                         page_set_storage_key(PFN_PHYS(pfn),
 867                                              PAGE_DEFAULT_KEY, 0);
 868         }
 869
 870         psw_set_key(PAGE_DEFAULT_KEY);
 871
 872         free_bootmem_with_active_regions(0, max_pfn);
 873
 874         /*
 875          * Reserve memory used for lowcore/command line/kernel image.
 876          */
 877         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 878         reserve_bootmem((unsigned long)_stext,
 879                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 880                         BOOTMEM_DEFAULT);
 881         /*
 882          * Reserve the bootmem bitmap itself as well. We do this in two
 883          * steps (first step was init_bootmem()) because this catches
 884          * the (very unlikely) case of us accidentally initializing the
 885          * bootmem allocator with an invalid RAM area.
 886          */
 887         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 888                         BOOTMEM_DEFAULT);
 889
 890 #ifdef CONFIG_CRASH_DUMP
 891         if (crashk_res.start)
 892                 reserve_bootmem(crashk_res.start,
 893                                 crashk_res.end - crashk_res.start + 1,
 894                                 BOOTMEM_DEFAULT);
 895         if (is_kdump_kernel())
 896                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
 897                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
 898 #endif
 899 #ifdef CONFIG_BLK_DEV_INITRD
 900         if (INITRD_START && INITRD_SIZE) {
 901                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 902                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 903                                         BOOTMEM_DEFAULT);
 904                         initrd_start = INITRD_START;
 905                         initrd_end = initrd_start + INITRD_SIZE;
 906                 } else {
 907                         pr_err("initrd extends beyond end of "
 908                                "memory (0x%08lx > 0x%08lx) "
 909                                "disabling initrd\n",
 910                                initrd_start + INITRD_SIZE, memory_end);
 911                         initrd_start = initrd_end = 0;
 912                 }
 913         }
 914 #endif
 915 }
 916
 917 /*
 918  * Setup hardware capabilities.
 919  */
 920 static void __init setup_hwcaps(void)
 921 {
 922         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 923         struct cpuid cpu_id;
 924         int i;
 925
 926         /*
 927          * The store facility list bits numbers as found in the principles
 928          * of operation are numbered with bit 1UL<<31 as number 0 to
 929          * bit 1UL<<0 as number 31.
 930          *   Bit 0: instructions named N3, "backported" to esa-mode
 931          *   Bit 2: z/Architecture mode is active
 932          *   Bit 7: the store-facility-list-extended facility is installed
 933          *   Bit 17: the message-security assist is installed
 934          *   Bit 19: the long-displacement facility is installed
 935          *   Bit 21: the extended-immediate facility is installed
 936          *   Bit 22: extended-translation facility 3 is installed
 937          *   Bit 30: extended-translation facility 3 enhancement facility
 938          * These get translated to:
 939          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 940          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 941          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 942          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 943          */
 944         for (i = 0; i < 6; i++)
 945                 if (test_facility(stfl_bits[i]))
 946                         elf_hwcap |= 1UL << i;
 947
 948         if (test_facility(22) && test_facility(30))
 949                 elf_hwcap |= HWCAP_S390_ETF3EH;
 950
 951         /*
 952          * Check for additional facilities with store-facility-list-extended.
 953          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 954          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 955          * as stored by stfl, bits 32-xxx contain additional facilities.
 956          * How many facility words are stored depends on the number of
 957          * doublewords passed to the instruction. The additional facilities
 958          * are:
 959          *   Bit 42: decimal floating point facility is installed
 960          *   Bit 44: perform floating point operation facility is installed
 961          * translated to:
 962          *   HWCAP_S390_DFP bit 6 (42 && 44).
 963          */
 964         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 965                 elf_hwcap |= HWCAP_S390_DFP;
 966
 967         /*
 968          * Huge page support HWCAP_S390_HPAGE is bit 7.
 969          */
 970         if (MACHINE_HAS_HPAGE)
 971                 elf_hwcap |= HWCAP_S390_HPAGE;
 972
 973         /*
 974          * 64-bit register support for 31-bit processes
 975          * HWCAP_S390_HIGH_GPRS is bit 9.
 976          */
 977         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 978
 979         get_cpu_id(&cpu_id);
 980         switch (cpu_id.machine) {
 981         case 0x9672:
 982 #if !defined(CONFIG_64BIT)
 983         default:        /* Use "g5" as default for 31 bit kernels. */
 984 #endif
 985                 strcpy(elf_platform, "g5");
 986                 break;
 987         case 0x2064:
 988         case 0x2066:
 989 #if defined(CONFIG_64BIT)
 990         default:        /* Use "z900" as default for 64 bit kernels. */
 991 #endif
 992                 strcpy(elf_platform, "z900");
 993                 break;
 994         case 0x2084:
 995         case 0x2086:
 996                 strcpy(elf_platform, "z990");
 997                 break;
 998         case 0x2094:
 999         case 0x2096:
1000                 strcpy(elf_platform, "z9-109");
1001                 break;
1002         case 0x2097:
1003         case 0x2098:
1004                 strcpy(elf_platform, "z10");
1005                 break;
1006         case 0x2817:
1007         case 0x2818:
1008                 strcpy(elf_platform, "z196");
1009                 break;
1010         }
1011 }
1012
1013 /*
1014  * Setup function called from init/main.c just after the banner
1015  * was printed.
1016  */
1017
1018 void __init setup_arch(char **cmdline_p)
1019 {
1020         /*
1021          * print what head.S has found out about the machine
1022          */
1023 #ifndef CONFIG_64BIT
1024         if (MACHINE_IS_VM)
1025                 pr_info("Linux is running as a z/VM "
1026                         "guest operating system in 31-bit mode\n");
1027         else if (MACHINE_IS_LPAR)
1028                 pr_info("Linux is running natively in 31-bit mode\n");
1029         if (MACHINE_HAS_IEEE)
1030                 pr_info("The hardware system has IEEE compatible "
1031                         "floating point units\n");
1032         else
1033                 pr_info("The hardware system has no IEEE compatible "
1034                         "floating point units\n");
1035 #else /* CONFIG_64BIT */
1036         if (MACHINE_IS_VM)
1037                 pr_info("Linux is running as a z/VM "
1038                         "guest operating system in 64-bit mode\n");
1039         else if (MACHINE_IS_KVM)
1040                 pr_info("Linux is running under KVM in 64-bit mode\n");
1041         else if (MACHINE_IS_LPAR)
1042                 pr_info("Linux is running natively in 64-bit mode\n");
1043 #endif /* CONFIG_64BIT */
1044
1045         /* Have one command line that is parsed and saved in /proc/cmdline */
1046         /* boot_command_line has been already set up in early.c */
1047         *cmdline_p = boot_command_line;
1048
1049         ROOT_DEV = Root_RAM0;
1050
1051         init_mm.start_code = PAGE_OFFSET;
1052         init_mm.end_code = (unsigned long) &_etext;
1053         init_mm.end_data = (unsigned long) &_edata;
1054         init_mm.brk = (unsigned long) &_end;
1055
1056         if (MACHINE_HAS_MVCOS)
1057                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1058         else
1059                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1060
1061         parse_early_param();
1062
1063         os_info_init();
1064         setup_ipl();
1065         setup_memory_end();
1066         setup_addressing_mode();
1067         reserve_oldmem();
1068         reserve_crashkernel();
1069         setup_memory();
1070         setup_resources();
1071         setup_vmcoreinfo();
1072         setup_lowcore();
1073
1074         cpu_init();
1075         s390_init_cpu_topology();
1076
1077         /*
1078          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1079          */
1080         setup_hwcaps();
1081
1082         /*
1083          * Create kernel page tables and switch to virtual addressing.
1084          */
1085         paging_init();
1086
1087         /* Setup default console */
1088         conmode_default();
1089         set_preferred_console();
1090
1091         /* Setup zfcpdump support */
1092         setup_zfcpdump(console_devno);
1093 }