arch/s390/kernel/setup.c

   1 /*
   2  *  S390 version
   3  *    Copyright IBM Corp. 1999, 2012
   4  *    Author(s): Hartmut Penner (hp@de.ibm.com),
   5  *               Martin Schwidefsky (schwidefsky@de.ibm.com)
   6  *
   7  *  Derived from "arch/i386/kernel/setup.c"
   8  *    Copyright (C) 1995, Linus Torvalds
   9  */
  10
  11 /*
  12  * This file handles the architecture-dependent parts of initialization
  13  */
  14
  15 #define KMSG_COMPONENT "setup"
  16 #define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
  17
  18 #include <linux/errno.h>
  19 #include <linux/module.h>
  20 #include <linux/sched.h>
  21 #include <linux/kernel.h>
  22 #include <linux/memblock.h>
  23 #include <linux/mm.h>
  24 #include <linux/stddef.h>
  25 #include <linux/unistd.h>
  26 #include <linux/ptrace.h>
  27 #include <linux/user.h>
  28 #include <linux/tty.h>
  29 #include <linux/ioport.h>
  30 #include <linux/delay.h>
  31 #include <linux/init.h>
  32 #include <linux/initrd.h>
  33 #include <linux/bootmem.h>
  34 #include <linux/root_dev.h>
  35 #include <linux/console.h>
  36 #include <linux/kernel_stat.h>
  37 #include <linux/device.h>
  38 #include <linux/notifier.h>
  39 #include <linux/pfn.h>
  40 #include <linux/ctype.h>
  41 #include <linux/reboot.h>
  42 #include <linux/topology.h>
  43 #include <linux/ftrace.h>
  44 #include <linux/kexec.h>
  45 #include <linux/crash_dump.h>
  46 #include <linux/memory.h>
  47 #include <linux/compat.h>
  48
  49 #include <asm/ipl.h>
  50 #include <asm/uaccess.h>
  51 #include <asm/facility.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/kvm_virtio.h>
  62 #include <asm/diag.h>
  63 #include <asm/os_info.h>
  64 #include "entry.h"
  65
  66 long psw_kernel_bits    = PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_ASC_PRIMARY |
  67                           PSW_MASK_EA | PSW_MASK_BA;
  68 long psw_user_bits      = PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT |
  69                           PSW_DEFAULT_KEY | PSW_MASK_BASE | PSW_MASK_MCHECK |
  70                           PSW_MASK_PSTATE | PSW_ASC_HOME;
  71
  72 /*
  73  * User copy operations.
  74  */
  75 struct uaccess_ops uaccess;
  76 EXPORT_SYMBOL(uaccess);
  77
  78 /*
  79  * Machine setup..
  80  */
  81 unsigned int console_mode = 0;
  82 EXPORT_SYMBOL(console_mode);
  83
  84 unsigned int console_devno = -1;
  85 EXPORT_SYMBOL(console_devno);
  86
  87 unsigned int console_irq = -1;
  88 EXPORT_SYMBOL(console_irq);
  89
  90 unsigned long elf_hwcap = 0;
  91 char elf_platform[ELF_PLATFORM_SIZE];
  92
  93 struct mem_chunk __initdata memory_chunk[MEMORY_CHUNKS];
  94
  95 int __initdata memory_end_set;
  96 unsigned long __initdata memory_end;
  97
  98 unsigned long VMALLOC_START;
  99 EXPORT_SYMBOL(VMALLOC_START);
 100
 101 unsigned long VMALLOC_END;
 102 EXPORT_SYMBOL(VMALLOC_END);
 103
 104 struct page *vmemmap;
 105 EXPORT_SYMBOL(vmemmap);
 106
 107 /* An array with a pointer to the lowcore of every CPU. */
 108 struct _lowcore *lowcore_ptr[NR_CPUS];
 109 EXPORT_SYMBOL(lowcore_ptr);
 110
 111 /*
 112  * This is set up by the setup-routine at boot-time
 113  * for S390 need to find out, what we have to setup
 114  * using address 0x10400 ...
 115  */
 116
 117 #include <asm/setup.h>
 118
 119 /*
 120  * condev= and conmode= setup parameter.
 121  */
 122
 123 static int __init condev_setup(char *str)
 124 {
 125         int vdev;
 126
 127         vdev = simple_strtoul(str, &str, 0);
 128         if (vdev >= 0 && vdev < 65536) {
 129                 console_devno = vdev;
 130                 console_irq = -1;
 131         }
 132         return 1;
 133 }
 134
 135 __setup("condev=", condev_setup);
 136
 137 static void __init set_preferred_console(void)
 138 {
 139         if (MACHINE_IS_KVM)
 140                 add_preferred_console("hvc", 0, NULL);
 141         else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 142                 add_preferred_console("ttyS", 0, NULL);
 143         else if (CONSOLE_IS_3270)
 144                 add_preferred_console("tty3270", 0, NULL);
 145 }
 146
 147 static int __init conmode_setup(char *str)
 148 {
 149 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 150         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 151                 SET_CONSOLE_SCLP;
 152 #endif
 153 #if defined(CONFIG_TN3215_CONSOLE)
 154         if (strncmp(str, "3215", 5) == 0)
 155                 SET_CONSOLE_3215;
 156 #endif
 157 #if defined(CONFIG_TN3270_CONSOLE)
 158         if (strncmp(str, "3270", 5) == 0)
 159                 SET_CONSOLE_3270;
 160 #endif
 161         set_preferred_console();
 162         return 1;
 163 }
 164
 165 __setup("conmode=", conmode_setup);
 166
 167 static void __init conmode_default(void)
 168 {
 169         char query_buffer[1024];
 170         char *ptr;
 171
 172         if (MACHINE_IS_VM) {
 173                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 174                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 175                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 176                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 177                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 178                 ptr = strstr(query_buffer, "CONMODE");
 179                 /*
 180                  * Set the conmode to 3215 so that the device recognition
 181                  * will set the cu_type of the console to 3215. If the
 182                  * conmode is 3270 and we don't set it back then both
 183                  * 3215 and the 3270 driver will try to access the console
 184                  * device (3215 as console and 3270 as normal tty).
 185                  */
 186                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 187                 if (ptr == NULL) {
 188 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 189                         SET_CONSOLE_SCLP;
 190 #endif
 191                         return;
 192                 }
 193                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 194 #if defined(CONFIG_TN3270_CONSOLE)
 195                         SET_CONSOLE_3270;
 196 #elif defined(CONFIG_TN3215_CONSOLE)
 197                         SET_CONSOLE_3215;
 198 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 199                         SET_CONSOLE_SCLP;
 200 #endif
 201                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 202 #if defined(CONFIG_TN3215_CONSOLE)
 203                         SET_CONSOLE_3215;
 204 #elif defined(CONFIG_TN3270_CONSOLE)
 205                         SET_CONSOLE_3270;
 206 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 207                         SET_CONSOLE_SCLP;
 208 #endif
 209                 }
 210         } else {
 211 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 212                 SET_CONSOLE_SCLP;
 213 #endif
 214         }
 215 }
 216
 217 #ifdef CONFIG_ZFCPDUMP
 218 static void __init setup_zfcpdump(unsigned int console_devno)
 219 {
 220         static char str[41];
 221
 222         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 223                 return;
 224         if (OLDMEM_BASE)
 225                 return;
 226         if (console_devno != -1)
 227                 sprintf(str, " cio_ignore=all,!0.0.%04x,!0.0.%04x",
 228                         ipl_info.data.fcp.dev_id.devno, console_devno);
 229         else
 230                 sprintf(str, " cio_ignore=all,!0.0.%04x",
 231                         ipl_info.data.fcp.dev_id.devno);
 232         strcat(boot_command_line, str);
 233         console_loglevel = 2;
 234 }
 235 #else
 236 static inline void setup_zfcpdump(unsigned int console_devno) {}
 237 #endif /* CONFIG_ZFCPDUMP */
 238
 239  /*
 240  * Reboot, halt and power_off stubs. They just call _machine_restart,
 241  * _machine_halt or _machine_power_off.
 242  */
 243
 244 void machine_restart(char *command)
 245 {
 246         if ((!in_interrupt() && !in_atomic()) || oops_in_progress)
 247                 /*
 248                  * Only unblank the console if we are called in enabled
 249                  * context or a bust_spinlocks cleared the way for us.
 250                  */
 251                 console_unblank();
 252         _machine_restart(command);
 253 }
 254
 255 void machine_halt(void)
 256 {
 257         if (!in_interrupt() || oops_in_progress)
 258                 /*
 259                  * Only unblank the console if we are called in enabled
 260                  * context or a bust_spinlocks cleared the way for us.
 261                  */
 262                 console_unblank();
 263         _machine_halt();
 264 }
 265
 266 void machine_power_off(void)
 267 {
 268         if (!in_interrupt() || oops_in_progress)
 269                 /*
 270                  * Only unblank the console if we are called in enabled
 271                  * context or a bust_spinlocks cleared the way for us.
 272                  */
 273                 console_unblank();
 274         _machine_power_off();
 275 }
 276
 277 /*
 278  * Dummy power off function.
 279  */
 280 void (*pm_power_off)(void) = machine_power_off;
 281
 282 static int __init early_parse_mem(char *p)
 283 {
 284         memory_end = memparse(p, &p);
 285         memory_end_set = 1;
 286         return 0;
 287 }
 288 early_param("mem", early_parse_mem);
 289
 290 static int __init parse_vmalloc(char *arg)
 291 {
 292         if (!arg)
 293                 return -EINVAL;
 294         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 295         return 0;
 296 }
 297 early_param("vmalloc", parse_vmalloc);
 298
 299 unsigned int user_mode = HOME_SPACE_MODE;
 300 EXPORT_SYMBOL_GPL(user_mode);
 301
 302 static int set_amode_primary(void)
 303 {
 304         psw_kernel_bits = (psw_kernel_bits & ~PSW_MASK_ASC) | PSW_ASC_HOME;
 305         psw_user_bits = (psw_user_bits & ~PSW_MASK_ASC) | PSW_ASC_PRIMARY;
 306 #ifdef CONFIG_COMPAT
 307         psw32_user_bits =
 308                 (psw32_user_bits & ~PSW32_MASK_ASC) | PSW32_ASC_PRIMARY;
 309 #endif
 310
 311         if (MACHINE_HAS_MVCOS) {
 312                 memcpy(&uaccess, &uaccess_mvcos_switch, sizeof(uaccess));
 313                 return 1;
 314         } else {
 315                 memcpy(&uaccess, &uaccess_pt, sizeof(uaccess));
 316                 return 0;
 317         }
 318 }
 319
 320 /*
 321  * Switch kernel/user addressing modes?
 322  */
 323 static int __init early_parse_switch_amode(char *p)
 324 {
 325         user_mode = PRIMARY_SPACE_MODE;
 326         return 0;
 327 }
 328 early_param("switch_amode", early_parse_switch_amode);
 329
 330 static int __init early_parse_user_mode(char *p)
 331 {
 332         if (p && strcmp(p, "primary") == 0)
 333                 user_mode = PRIMARY_SPACE_MODE;
 334         else if (!p || strcmp(p, "home") == 0)
 335                 user_mode = HOME_SPACE_MODE;
 336         else
 337                 return 1;
 338         return 0;
 339 }
 340 early_param("user_mode", early_parse_user_mode);
 341
 342 static void setup_addressing_mode(void)
 343 {
 344         if (user_mode == PRIMARY_SPACE_MODE) {
 345                 if (set_amode_primary())
 346                         pr_info("Address spaces switched, "
 347                                 "mvcos available\n");
 348                 else
 349                         pr_info("Address spaces switched, "
 350                                 "mvcos not available\n");
 351         }
 352 }
 353
 354 void *restart_stack __attribute__((__section__(".data")));
 355
 356 static void __init setup_lowcore(void)
 357 {
 358         struct _lowcore *lc;
 359
 360         /*
 361          * Setup lowcore for boot cpu
 362          */
 363         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 364         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 365         lc->restart_psw.mask = psw_kernel_bits;
 366         lc->restart_psw.addr =
 367                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 368         lc->external_new_psw.mask = psw_kernel_bits |
 369                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 370         lc->external_new_psw.addr =
 371                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 372         lc->svc_new_psw.mask = psw_kernel_bits |
 373                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 374         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 375         lc->program_new_psw.mask = psw_kernel_bits |
 376                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 377         lc->program_new_psw.addr =
 378                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 379         lc->mcck_new_psw.mask = psw_kernel_bits;
 380         lc->mcck_new_psw.addr =
 381                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 382         lc->io_new_psw.mask = psw_kernel_bits |
 383                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 384         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 385         lc->clock_comparator = -1ULL;
 386         lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
 387         lc->async_stack = (unsigned long)
 388                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
 389         lc->panic_stack = (unsigned long)
 390                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
 391         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 392         lc->thread_info = (unsigned long) &init_thread_union;
 393         lc->machine_flags = S390_lowcore.machine_flags;
 394         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 395         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 396                MAX_FACILITY_BIT/8);
 397 #ifndef CONFIG_64BIT
 398         if (MACHINE_HAS_IEEE) {
 399                 lc->extended_save_area_addr = (__u32)
 400                         __alloc_bootmem_low(PAGE_SIZE, PAGE_SIZE, 0);
 401                 /* enable extended save area */
 402                 __ctl_set_bit(14, 29);
 403         }
 404 #else
 405         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 406 #endif
 407         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 408         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 409         lc->exit_timer = S390_lowcore.exit_timer;
 410         lc->user_timer = S390_lowcore.user_timer;
 411         lc->system_timer = S390_lowcore.system_timer;
 412         lc->steal_timer = S390_lowcore.steal_timer;
 413         lc->last_update_timer = S390_lowcore.last_update_timer;
 414         lc->last_update_clock = S390_lowcore.last_update_clock;
 415         lc->ftrace_func = S390_lowcore.ftrace_func;
 416
 417         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 418         restart_stack += ASYNC_SIZE;
 419
 420         /*
 421          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 422          * restart data to the absolute zero lowcore. This is necesary if
 423          * PSW restart is done on an offline CPU that has lowcore zero.
 424          */
 425         lc->restart_stack = (unsigned long) restart_stack;
 426         lc->restart_fn = (unsigned long) do_restart;
 427         lc->restart_data = 0;
 428         lc->restart_source = -1UL;
 429
 430         /* Setup absolute zero lowcore */
 431         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 432         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 433         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 434         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 435         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 436
 437         set_prefix((u32)(unsigned long) lc);
 438         lowcore_ptr[0] = lc;
 439 }
 440
 441 static struct resource code_resource = {
 442         .name  = "Kernel code",
 443         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 444 };
 445
 446 static struct resource data_resource = {
 447         .name = "Kernel data",
 448         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 449 };
 450
 451 static struct resource bss_resource = {
 452         .name = "Kernel bss",
 453         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 454 };
 455
 456 static struct resource __initdata *standard_resources[] = {
 457         &code_resource,
 458         &data_resource,
 459         &bss_resource,
 460 };
 461
 462 static void __init setup_resources(void)
 463 {
 464         struct resource *res, *std_res, *sub_res;
 465         int i, j;
 466
 467         code_resource.start = (unsigned long) &_text;
 468         code_resource.end = (unsigned long) &_etext - 1;
 469         data_resource.start = (unsigned long) &_etext;
 470         data_resource.end = (unsigned long) &_edata - 1;
 471         bss_resource.start = (unsigned long) &__bss_start;
 472         bss_resource.end = (unsigned long) &__bss_stop - 1;
 473
 474         for (i = 0; i < MEMORY_CHUNKS; i++) {
 475                 if (!memory_chunk[i].size)
 476                         continue;
 477                 if (memory_chunk[i].type == CHUNK_OLDMEM ||
 478                     memory_chunk[i].type == CHUNK_CRASHK)
 479                         continue;
 480                 res = alloc_bootmem_low(sizeof(*res));
 481                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 482                 switch (memory_chunk[i].type) {
 483                 case CHUNK_READ_WRITE:
 484                 case CHUNK_CRASHK:
 485                         res->name = "System RAM";
 486                         break;
 487                 case CHUNK_READ_ONLY:
 488                         res->name = "System ROM";
 489                         res->flags |= IORESOURCE_READONLY;
 490                         break;
 491                 default:
 492                         res->name = "reserved";
 493                 }
 494                 res->start = memory_chunk[i].addr;
 495                 res->end = res->start + memory_chunk[i].size - 1;
 496                 request_resource(&iomem_resource, res);
 497
 498                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 499                         std_res = standard_resources[j];
 500                         if (std_res->start < res->start ||
 501                             std_res->start > res->end)
 502                                 continue;
 503                         if (std_res->end > res->end) {
 504                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 505                                 *sub_res = *std_res;
 506                                 sub_res->end = res->end;
 507                                 std_res->start = res->end + 1;
 508                                 request_resource(res, sub_res);
 509                         } else {
 510                                 request_resource(res, std_res);
 511                         }
 512                 }
 513         }
 514 }
 515
 516 unsigned long real_memory_size;
 517 EXPORT_SYMBOL_GPL(real_memory_size);
 518
 519 static void __init setup_memory_end(void)
 520 {
 521         unsigned long vmax, vmalloc_size, tmp;
 522         int i;
 523
 524
 525 #ifdef CONFIG_ZFCPDUMP
 526         if (ipl_info.type == IPL_TYPE_FCP_DUMP && !OLDMEM_BASE) {
 527                 memory_end = ZFCPDUMP_HSA_SIZE;
 528                 memory_end_set = 1;
 529         }
 530 #endif
 531         real_memory_size = 0;
 532         memory_end &= PAGE_MASK;
 533
 534         /*
 535          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 536          * extra checks that HOLES_IN_ZONE would require.
 537          */
 538         for (i = 0; i < MEMORY_CHUNKS; i++) {
 539                 unsigned long start, end;
 540                 struct mem_chunk *chunk;
 541                 unsigned long align;
 542
 543                 chunk = &memory_chunk[i];
 544                 align = 1UL << (MAX_ORDER + PAGE_SHIFT - 1);
 545                 start = (chunk->addr + align - 1) & ~(align - 1);
 546                 end = (chunk->addr + chunk->size) & ~(align - 1);
 547                 if (start >= end)
 548                         memset(chunk, 0, sizeof(*chunk));
 549                 else {
 550                         chunk->addr = start;
 551                         chunk->size = end - start;
 552                 }
 553                 real_memory_size = max(real_memory_size,
 554                                        chunk->addr + chunk->size);
 555         }
 556
 557         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 558 #ifdef CONFIG_64BIT
 559         vmalloc_size = VMALLOC_END ?: 128UL << 30;
 560         tmp = (memory_end ?: real_memory_size) / PAGE_SIZE;
 561         tmp = tmp * (sizeof(struct page) + PAGE_SIZE) + vmalloc_size;
 562         if (tmp <= (1UL << 42))
 563                 vmax = 1UL << 42;       /* 3-level kernel page table */
 564         else
 565                 vmax = 1UL << 53;       /* 4-level kernel page table */
 566 #else
 567         vmalloc_size = VMALLOC_END ?: 96UL << 20;
 568         vmax = 1UL << 31;               /* 2-level kernel page table */
 569 #endif
 570         /* vmalloc area is at the end of the kernel address space. */
 571         VMALLOC_END = vmax;
 572         VMALLOC_START = vmax - vmalloc_size;
 573
 574         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 575         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 576         tmp = VMALLOC_START - tmp * sizeof(struct page);
 577         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 578         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 579         vmemmap = (struct page *) tmp;
 580
 581         /* Take care that memory_end is set and <= vmemmap */
 582         memory_end = min(memory_end ?: real_memory_size, tmp);
 583
 584         /* Fixup memory chunk array to fit into 0..memory_end */
 585         for (i = 0; i < MEMORY_CHUNKS; i++) {
 586                 struct mem_chunk *chunk = &memory_chunk[i];
 587
 588                 if (chunk->addr >= memory_end) {
 589                         memset(chunk, 0, sizeof(*chunk));
 590                         continue;
 591                 }
 592                 if (chunk->addr + chunk->size > memory_end)
 593                         chunk->size = memory_end - chunk->addr;
 594         }
 595 }
 596
 597 static void __init setup_vmcoreinfo(void)
 598 {
 599 #ifdef CONFIG_KEXEC
 600         mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
 601 #endif
 602 }
 603
 604 #ifdef CONFIG_CRASH_DUMP
 605
 606 /*
 607  * Find suitable location for crashkernel memory
 608  */
 609 static unsigned long __init find_crash_base(unsigned long crash_size,
 610                                             char **msg)
 611 {
 612         unsigned long crash_base;
 613         struct mem_chunk *chunk;
 614         int i;
 615
 616         if (memory_chunk[0].size < crash_size) {
 617                 *msg = "first memory chunk must be at least crashkernel size";
 618                 return 0;
 619         }
 620         if (OLDMEM_BASE && crash_size == OLDMEM_SIZE)
 621                 return OLDMEM_BASE;
 622
 623         for (i = MEMORY_CHUNKS - 1; i >= 0; i--) {
 624                 chunk = &memory_chunk[i];
 625                 if (chunk->size == 0)
 626                         continue;
 627                 if (chunk->type != CHUNK_READ_WRITE)
 628                         continue;
 629                 if (chunk->size < crash_size)
 630                         continue;
 631                 crash_base = (chunk->addr + chunk->size) - crash_size;
 632                 if (crash_base < crash_size)
 633                         continue;
 634                 if (crash_base < ZFCPDUMP_HSA_SIZE_MAX)
 635                         continue;
 636                 if (crash_base < (unsigned long) INITRD_START + INITRD_SIZE)
 637                         continue;
 638                 return crash_base;
 639         }
 640         *msg = "no suitable area found";
 641         return 0;
 642 }
 643
 644 /*
 645  * Check if crash_base and crash_size is valid
 646  */
 647 static int __init verify_crash_base(unsigned long crash_base,
 648                                     unsigned long crash_size,
 649                                     char **msg)
 650 {
 651         struct mem_chunk *chunk;
 652         int i;
 653
 654         /*
 655          * Because we do the swap to zero, we must have at least 'crash_size'
 656          * bytes free space before crash_base
 657          */
 658         if (crash_size > crash_base) {
 659                 *msg = "crashkernel offset must be greater than size";
 660                 return -EINVAL;
 661         }
 662
 663         /* First memory chunk must be at least crash_size */
 664         if (memory_chunk[0].size < crash_size) {
 665                 *msg = "first memory chunk must be at least crashkernel size";
 666                 return -EINVAL;
 667         }
 668         /* Check if we fit into the respective memory chunk */
 669         for (i = 0; i < MEMORY_CHUNKS; i++) {
 670                 chunk = &memory_chunk[i];
 671                 if (chunk->size == 0)
 672                         continue;
 673                 if (crash_base < chunk->addr)
 674                         continue;
 675                 if (crash_base >= chunk->addr + chunk->size)
 676                         continue;
 677                 /* we have found the memory chunk */
 678                 if (crash_base + crash_size > chunk->addr + chunk->size) {
 679                         *msg = "selected memory chunk is too small for "
 680                                 "crashkernel memory";
 681                         return -EINVAL;
 682                 }
 683                 return 0;
 684         }
 685         *msg = "invalid memory range specified";
 686         return -EINVAL;
 687 }
 688
 689 /*
 690  * Reserve kdump memory by creating a memory hole in the mem_chunk array
 691  */
 692 static void __init reserve_kdump_bootmem(unsigned long addr, unsigned long size,
 693                                          int type)
 694 {
 695         create_mem_hole(memory_chunk, addr, size, type);
 696 }
 697
 698 /*
 699  * When kdump is enabled, we have to ensure that no memory from
 700  * the area [0 - crashkernel memory size] and
 701  * [crashk_res.start - crashk_res.end] is set offline.
 702  */
 703 static int kdump_mem_notifier(struct notifier_block *nb,
 704                               unsigned long action, void *data)
 705 {
 706         struct memory_notify *arg = data;
 707
 708         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 709                 return NOTIFY_BAD;
 710         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 711                 return NOTIFY_OK;
 712         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 713                 return NOTIFY_OK;
 714         return NOTIFY_BAD;
 715 }
 716
 717 static struct notifier_block kdump_mem_nb = {
 718         .notifier_call = kdump_mem_notifier,
 719 };
 720
 721 #endif
 722
 723 /*
 724  * Make sure that oldmem, where the dump is stored, is protected
 725  */
 726 static void reserve_oldmem(void)
 727 {
 728 #ifdef CONFIG_CRASH_DUMP
 729         if (!OLDMEM_BASE)
 730                 return;
 731
 732         reserve_kdump_bootmem(OLDMEM_BASE, OLDMEM_SIZE, CHUNK_OLDMEM);
 733         reserve_kdump_bootmem(OLDMEM_SIZE, memory_end - OLDMEM_SIZE,
 734                               CHUNK_OLDMEM);
 735         if (OLDMEM_BASE + OLDMEM_SIZE == real_memory_size)
 736                 saved_max_pfn = PFN_DOWN(OLDMEM_BASE) - 1;
 737         else
 738                 saved_max_pfn = PFN_DOWN(real_memory_size) - 1;
 739 #endif
 740 }
 741
 742 /*
 743  * Reserve memory for kdump kernel to be loaded with kexec
 744  */
 745 static void __init reserve_crashkernel(void)
 746 {
 747 #ifdef CONFIG_CRASH_DUMP
 748         unsigned long long crash_base, crash_size;
 749         char *msg = NULL;
 750         int rc;
 751
 752         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 753                                &crash_base);
 754         if (rc || crash_size == 0)
 755                 return;
 756         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 757         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 758         if (register_memory_notifier(&kdump_mem_nb))
 759                 return;
 760         if (!crash_base)
 761                 crash_base = find_crash_base(crash_size, &msg);
 762         if (!crash_base) {
 763                 pr_info("crashkernel reservation failed: %s\n", msg);
 764                 unregister_memory_notifier(&kdump_mem_nb);
 765                 return;
 766         }
 767         if (verify_crash_base(crash_base, crash_size, &msg)) {
 768                 pr_info("crashkernel reservation failed: %s\n", msg);
 769                 unregister_memory_notifier(&kdump_mem_nb);
 770                 return;
 771         }
 772         if (!OLDMEM_BASE && MACHINE_IS_VM)
 773                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 774         crashk_res.start = crash_base;
 775         crashk_res.end = crash_base + crash_size - 1;
 776         insert_resource(&iomem_resource, &crashk_res);
 777         reserve_kdump_bootmem(crash_base, crash_size, CHUNK_CRASHK);
 778         pr_info("Reserving %lluMB of memory at %lluMB "
 779                 "for crashkernel (System RAM: %luMB)\n",
 780                 crash_size >> 20, crash_base >> 20, memory_end >> 20);
 781         os_info_crashkernel_add(crash_base, crash_size);
 782 #endif
 783 }
 784
 785 static void __init setup_memory(void)
 786 {
 787         unsigned long bootmap_size;
 788         unsigned long start_pfn, end_pfn;
 789         int i;
 790
 791         /*
 792          * partially used pages are not usable - thus
 793          * we are rounding upwards:
 794          */
 795         start_pfn = PFN_UP(__pa(&_end));
 796         end_pfn = max_pfn = PFN_DOWN(memory_end);
 797
 798 #ifdef CONFIG_BLK_DEV_INITRD
 799         /*
 800          * Move the initrd in case the bitmap of the bootmem allocater
 801          * would overwrite it.
 802          */
 803
 804         if (INITRD_START && INITRD_SIZE) {
 805                 unsigned long bmap_size;
 806                 unsigned long start;
 807
 808                 bmap_size = bootmem_bootmap_pages(end_pfn - start_pfn + 1);
 809                 bmap_size = PFN_PHYS(bmap_size);
 810
 811                 if (PFN_PHYS(start_pfn) + bmap_size > INITRD_START) {
 812                         start = PFN_PHYS(start_pfn) + bmap_size + PAGE_SIZE;
 813
 814 #ifdef CONFIG_CRASH_DUMP
 815                         if (OLDMEM_BASE) {
 816                                 /* Move initrd behind kdump oldmem */
 817                                 if (start + INITRD_SIZE > OLDMEM_BASE &&
 818                                     start < OLDMEM_BASE + OLDMEM_SIZE)
 819                                         start = OLDMEM_BASE + OLDMEM_SIZE;
 820                         }
 821 #endif
 822                         if (start + INITRD_SIZE > memory_end) {
 823                                 pr_err("initrd extends beyond end of "
 824                                        "memory (0x%08lx > 0x%08lx) "
 825                                        "disabling initrd\n",
 826                                        start + INITRD_SIZE, memory_end);
 827                                 INITRD_START = INITRD_SIZE = 0;
 828                         } else {
 829                                 pr_info("Moving initrd (0x%08lx -> "
 830                                         "0x%08lx, size: %ld)\n",
 831                                         INITRD_START, start, INITRD_SIZE);
 832                                 memmove((void *) start, (void *) INITRD_START,
 833                                         INITRD_SIZE);
 834                                 INITRD_START = start;
 835                         }
 836                 }
 837         }
 838 #endif
 839
 840         /*
 841          * Initialize the boot-time allocator
 842          */
 843         bootmap_size = init_bootmem(start_pfn, end_pfn);
 844
 845         /*
 846          * Register RAM areas with the bootmem allocator.
 847          */
 848
 849         for (i = 0; i < MEMORY_CHUNKS && memory_chunk[i].size > 0; i++) {
 850                 unsigned long start_chunk, end_chunk, pfn;
 851
 852                 if (memory_chunk[i].type != CHUNK_READ_WRITE &&
 853                     memory_chunk[i].type != CHUNK_CRASHK)
 854                         continue;
 855                 start_chunk = PFN_DOWN(memory_chunk[i].addr);
 856                 end_chunk = start_chunk + PFN_DOWN(memory_chunk[i].size);
 857                 end_chunk = min(end_chunk, end_pfn);
 858                 if (start_chunk >= end_chunk)
 859                         continue;
 860                 memblock_add_node(PFN_PHYS(start_chunk),
 861                                   PFN_PHYS(end_chunk - start_chunk), 0);
 862                 pfn = max(start_chunk, start_pfn);
 863                 for (; pfn < end_chunk; pfn++)
 864                         page_set_storage_key(PFN_PHYS(pfn),
 865                                              PAGE_DEFAULT_KEY, 0);
 866         }
 867
 868         psw_set_key(PAGE_DEFAULT_KEY);
 869
 870         free_bootmem_with_active_regions(0, max_pfn);
 871
 872         /*
 873          * Reserve memory used for lowcore/command line/kernel image.
 874          */
 875         reserve_bootmem(0, (unsigned long)_ehead, BOOTMEM_DEFAULT);
 876         reserve_bootmem((unsigned long)_stext,
 877                         PFN_PHYS(start_pfn) - (unsigned long)_stext,
 878                         BOOTMEM_DEFAULT);
 879         /*
 880          * Reserve the bootmem bitmap itself as well. We do this in two
 881          * steps (first step was init_bootmem()) because this catches
 882          * the (very unlikely) case of us accidentally initializing the
 883          * bootmem allocator with an invalid RAM area.
 884          */
 885         reserve_bootmem(start_pfn << PAGE_SHIFT, bootmap_size,
 886                         BOOTMEM_DEFAULT);
 887
 888 #ifdef CONFIG_CRASH_DUMP
 889         if (crashk_res.start)
 890                 reserve_bootmem(crashk_res.start,
 891                                 crashk_res.end - crashk_res.start + 1,
 892                                 BOOTMEM_DEFAULT);
 893         if (is_kdump_kernel())
 894                 reserve_bootmem(elfcorehdr_addr - OLDMEM_BASE,
 895                                 PAGE_ALIGN(elfcorehdr_size), BOOTMEM_DEFAULT);
 896 #endif
 897 #ifdef CONFIG_BLK_DEV_INITRD
 898         if (INITRD_START && INITRD_SIZE) {
 899                 if (INITRD_START + INITRD_SIZE <= memory_end) {
 900                         reserve_bootmem(INITRD_START, INITRD_SIZE,
 901                                         BOOTMEM_DEFAULT);
 902                         initrd_start = INITRD_START;
 903                         initrd_end = initrd_start + INITRD_SIZE;
 904                 } else {
 905                         pr_err("initrd extends beyond end of "
 906                                "memory (0x%08lx > 0x%08lx) "
 907                                "disabling initrd\n",
 908                                initrd_start + INITRD_SIZE, memory_end);
 909                         initrd_start = initrd_end = 0;
 910                 }
 911         }
 912 #endif
 913 }
 914
 915 /*
 916  * Setup hardware capabilities.
 917  */
 918 static void __init setup_hwcaps(void)
 919 {
 920         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 921         struct cpuid cpu_id;
 922         int i;
 923
 924         /*
 925          * The store facility list bits numbers as found in the principles
 926          * of operation are numbered with bit 1UL<<31 as number 0 to
 927          * bit 1UL<<0 as number 31.
 928          *   Bit 0: instructions named N3, "backported" to esa-mode
 929          *   Bit 2: z/Architecture mode is active
 930          *   Bit 7: the store-facility-list-extended facility is installed
 931          *   Bit 17: the message-security assist is installed
 932          *   Bit 19: the long-displacement facility is installed
 933          *   Bit 21: the extended-immediate facility is installed
 934          *   Bit 22: extended-translation facility 3 is installed
 935          *   Bit 30: extended-translation facility 3 enhancement facility
 936          * These get translated to:
 937          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 938          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 939          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 940          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 941          */
 942         for (i = 0; i < 6; i++)
 943                 if (test_facility(stfl_bits[i]))
 944                         elf_hwcap |= 1UL << i;
 945
 946         if (test_facility(22) && test_facility(30))
 947                 elf_hwcap |= HWCAP_S390_ETF3EH;
 948
 949         /*
 950          * Check for additional facilities with store-facility-list-extended.
 951          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 952          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 953          * as stored by stfl, bits 32-xxx contain additional facilities.
 954          * How many facility words are stored depends on the number of
 955          * doublewords passed to the instruction. The additional facilities
 956          * are:
 957          *   Bit 42: decimal floating point facility is installed
 958          *   Bit 44: perform floating point operation facility is installed
 959          * translated to:
 960          *   HWCAP_S390_DFP bit 6 (42 && 44).
 961          */
 962         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 963                 elf_hwcap |= HWCAP_S390_DFP;
 964
 965         /*
 966          * Huge page support HWCAP_S390_HPAGE is bit 7.
 967          */
 968         if (MACHINE_HAS_HPAGE)
 969                 elf_hwcap |= HWCAP_S390_HPAGE;
 970
 971         /*
 972          * 64-bit register support for 31-bit processes
 973          * HWCAP_S390_HIGH_GPRS is bit 9.
 974          */
 975         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 976
 977         get_cpu_id(&cpu_id);
 978         switch (cpu_id.machine) {
 979         case 0x9672:
 980 #if !defined(CONFIG_64BIT)
 981         default:        /* Use "g5" as default for 31 bit kernels. */
 982 #endif
 983                 strcpy(elf_platform, "g5");
 984                 break;
 985         case 0x2064:
 986         case 0x2066:
 987 #if defined(CONFIG_64BIT)
 988         default:        /* Use "z900" as default for 64 bit kernels. */
 989 #endif
 990                 strcpy(elf_platform, "z900");
 991                 break;
 992         case 0x2084:
 993         case 0x2086:
 994                 strcpy(elf_platform, "z990");
 995                 break;
 996         case 0x2094:
 997         case 0x2096:
 998                 strcpy(elf_platform, "z9-109");
 999                 break;
1000         case 0x2097:
1001         case 0x2098:
1002                 strcpy(elf_platform, "z10");
1003                 break;
1004         case 0x2817:
1005         case 0x2818:
1006                 strcpy(elf_platform, "z196");
1007                 break;
1008         }
1009 }
1010
1011 /*
1012  * Setup function called from init/main.c just after the banner
1013  * was printed.
1014  */
1015
1016 void __init setup_arch(char **cmdline_p)
1017 {
1018         /*
1019          * print what head.S has found out about the machine
1020          */
1021 #ifndef CONFIG_64BIT
1022         if (MACHINE_IS_VM)
1023                 pr_info("Linux is running as a z/VM "
1024                         "guest operating system in 31-bit mode\n");
1025         else if (MACHINE_IS_LPAR)
1026                 pr_info("Linux is running natively in 31-bit mode\n");
1027         if (MACHINE_HAS_IEEE)
1028                 pr_info("The hardware system has IEEE compatible "
1029                         "floating point units\n");
1030         else
1031                 pr_info("The hardware system has no IEEE compatible "
1032                         "floating point units\n");
1033 #else /* CONFIG_64BIT */
1034         if (MACHINE_IS_VM)
1035                 pr_info("Linux is running as a z/VM "
1036                         "guest operating system in 64-bit mode\n");
1037         else if (MACHINE_IS_KVM)
1038                 pr_info("Linux is running under KVM in 64-bit mode\n");
1039         else if (MACHINE_IS_LPAR)
1040                 pr_info("Linux is running natively in 64-bit mode\n");
1041 #endif /* CONFIG_64BIT */
1042
1043         /* Have one command line that is parsed and saved in /proc/cmdline */
1044         /* boot_command_line has been already set up in early.c */
1045         *cmdline_p = boot_command_line;
1046
1047         ROOT_DEV = Root_RAM0;
1048
1049         init_mm.start_code = PAGE_OFFSET;
1050         init_mm.end_code = (unsigned long) &_etext;
1051         init_mm.end_data = (unsigned long) &_edata;
1052         init_mm.brk = (unsigned long) &_end;
1053
1054         if (MACHINE_HAS_MVCOS)
1055                 memcpy(&uaccess, &uaccess_mvcos, sizeof(uaccess));
1056         else
1057                 memcpy(&uaccess, &uaccess_std, sizeof(uaccess));
1058
1059         parse_early_param();
1060
1061         os_info_init();
1062         setup_ipl();
1063         setup_memory_end();
1064         setup_addressing_mode();
1065         reserve_oldmem();
1066         reserve_crashkernel();
1067         setup_memory();
1068         setup_resources();
1069         setup_vmcoreinfo();
1070         setup_lowcore();
1071
1072         cpu_init();
1073         s390_init_cpu_topology();
1074
1075         /*
1076          * Setup capabilities (ELF_HWCAP & ELF_PLATFORM).
1077          */
1078         setup_hwcaps();
1079
1080         /*
1081          * Create kernel page tables and switch to virtual addressing.
1082          */
1083         paging_init();
1084
1085         /* Setup default console */
1086         conmode_default();
1087         set_preferred_console();
1088
1089         /* Setup zfcpdump support */
1090         setup_zfcpdump(console_devno);
1091 }