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/export.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/random.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/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/facility.h>
  51 #include <asm/smp.h>
  52 #include <asm/mmu_context.h>
  53 #include <asm/cpcmd.h>
  54 #include <asm/lowcore.h>
  55 #include <asm/irq.h>
  56 #include <asm/page.h>
  57 #include <asm/ptrace.h>
  58 #include <asm/sections.h>
  59 #include <asm/ebcdic.h>
  60 #include <asm/kvm_virtio.h>
  61 #include <asm/diag.h>
  62 #include <asm/os_info.h>
  63 #include <asm/sclp.h>
  64 #include <asm/sysinfo.h>
  65 #include "entry.h"
  66
  67 /*
  68  * Machine setup..
  69  */
  70 unsigned int console_mode = 0;
  71 EXPORT_SYMBOL(console_mode);
  72
  73 unsigned int console_devno = -1;
  74 EXPORT_SYMBOL(console_devno);
  75
  76 unsigned int console_irq = -1;
  77 EXPORT_SYMBOL(console_irq);
  78
  79 unsigned long elf_hwcap = 0;
  80 char elf_platform[ELF_PLATFORM_SIZE];
  81
  82 int __initdata memory_end_set;
  83 unsigned long __initdata memory_end;
  84 unsigned long __initdata max_physmem_end;
  85
  86 unsigned long VMALLOC_START;
  87 EXPORT_SYMBOL(VMALLOC_START);
  88
  89 unsigned long VMALLOC_END;
  90 EXPORT_SYMBOL(VMALLOC_END);
  91
  92 struct page *vmemmap;
  93 EXPORT_SYMBOL(vmemmap);
  94
  95 unsigned long MODULES_VADDR;
  96 unsigned long MODULES_END;
  97
  98 /* An array with a pointer to the lowcore of every CPU. */
  99 struct _lowcore *lowcore_ptr[NR_CPUS];
 100 EXPORT_SYMBOL(lowcore_ptr);
 101
 102 /*
 103  * This is set up by the setup-routine at boot-time
 104  * for S390 need to find out, what we have to setup
 105  * using address 0x10400 ...
 106  */
 107
 108 #include <asm/setup.h>
 109
 110 /*
 111  * condev= and conmode= setup parameter.
 112  */
 113
 114 static int __init condev_setup(char *str)
 115 {
 116         int vdev;
 117
 118         vdev = simple_strtoul(str, &str, 0);
 119         if (vdev >= 0 && vdev < 65536) {
 120                 console_devno = vdev;
 121                 console_irq = -1;
 122         }
 123         return 1;
 124 }
 125
 126 __setup("condev=", condev_setup);
 127
 128 static void __init set_preferred_console(void)
 129 {
 130         if (MACHINE_IS_KVM) {
 131                 if (sclp.has_vt220)
 132                         add_preferred_console("ttyS", 1, NULL);
 133                 else if (sclp.has_linemode)
 134                         add_preferred_console("ttyS", 0, NULL);
 135                 else
 136                         add_preferred_console("hvc", 0, NULL);
 137         } else if (CONSOLE_IS_3215 || CONSOLE_IS_SCLP)
 138                 add_preferred_console("ttyS", 0, NULL);
 139         else if (CONSOLE_IS_3270)
 140                 add_preferred_console("tty3270", 0, NULL);
 141 }
 142
 143 static int __init conmode_setup(char *str)
 144 {
 145 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 146         if (strncmp(str, "hwc", 4) == 0 || strncmp(str, "sclp", 5) == 0)
 147                 SET_CONSOLE_SCLP;
 148 #endif
 149 #if defined(CONFIG_TN3215_CONSOLE)
 150         if (strncmp(str, "3215", 5) == 0)
 151                 SET_CONSOLE_3215;
 152 #endif
 153 #if defined(CONFIG_TN3270_CONSOLE)
 154         if (strncmp(str, "3270", 5) == 0)
 155                 SET_CONSOLE_3270;
 156 #endif
 157         set_preferred_console();
 158         return 1;
 159 }
 160
 161 __setup("conmode=", conmode_setup);
 162
 163 static void __init conmode_default(void)
 164 {
 165         char query_buffer[1024];
 166         char *ptr;
 167
 168         if (MACHINE_IS_VM) {
 169                 cpcmd("QUERY CONSOLE", query_buffer, 1024, NULL);
 170                 console_devno = simple_strtoul(query_buffer + 5, NULL, 16);
 171                 ptr = strstr(query_buffer, "SUBCHANNEL =");
 172                 console_irq = simple_strtoul(ptr + 13, NULL, 16);
 173                 cpcmd("QUERY TERM", query_buffer, 1024, NULL);
 174                 ptr = strstr(query_buffer, "CONMODE");
 175                 /*
 176                  * Set the conmode to 3215 so that the device recognition
 177                  * will set the cu_type of the console to 3215. If the
 178                  * conmode is 3270 and we don't set it back then both
 179                  * 3215 and the 3270 driver will try to access the console
 180                  * device (3215 as console and 3270 as normal tty).
 181                  */
 182                 cpcmd("TERM CONMODE 3215", NULL, 0, NULL);
 183                 if (ptr == NULL) {
 184 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 185                         SET_CONSOLE_SCLP;
 186 #endif
 187                         return;
 188                 }
 189                 if (strncmp(ptr + 8, "3270", 4) == 0) {
 190 #if defined(CONFIG_TN3270_CONSOLE)
 191                         SET_CONSOLE_3270;
 192 #elif defined(CONFIG_TN3215_CONSOLE)
 193                         SET_CONSOLE_3215;
 194 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 195                         SET_CONSOLE_SCLP;
 196 #endif
 197                 } else if (strncmp(ptr + 8, "3215", 4) == 0) {
 198 #if defined(CONFIG_TN3215_CONSOLE)
 199                         SET_CONSOLE_3215;
 200 #elif defined(CONFIG_TN3270_CONSOLE)
 201                         SET_CONSOLE_3270;
 202 #elif defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 203                         SET_CONSOLE_SCLP;
 204 #endif
 205                 }
 206         } else {
 207 #if defined(CONFIG_SCLP_CONSOLE) || defined(CONFIG_SCLP_VT220_CONSOLE)
 208                 SET_CONSOLE_SCLP;
 209 #endif
 210         }
 211 }
 212
 213 #ifdef CONFIG_CRASH_DUMP
 214 static void __init setup_zfcpdump(void)
 215 {
 216         if (ipl_info.type != IPL_TYPE_FCP_DUMP)
 217                 return;
 218         if (OLDMEM_BASE)
 219                 return;
 220         strcat(boot_command_line, " cio_ignore=all,!ipldev,!condev");
 221         console_loglevel = 2;
 222 }
 223 #else
 224 static inline void setup_zfcpdump(void) {}
 225 #endif /* CONFIG_CRASH_DUMP */
 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 EXPORT_SYMBOL_GPL(pm_power_off);
 270
 271 static int __init early_parse_mem(char *p)
 272 {
 273         memory_end = memparse(p, &p);
 274         memory_end &= PAGE_MASK;
 275         memory_end_set = 1;
 276         return 0;
 277 }
 278 early_param("mem", early_parse_mem);
 279
 280 static int __init parse_vmalloc(char *arg)
 281 {
 282         if (!arg)
 283                 return -EINVAL;
 284         VMALLOC_END = (memparse(arg, &arg) + PAGE_SIZE - 1) & PAGE_MASK;
 285         return 0;
 286 }
 287 early_param("vmalloc", parse_vmalloc);
 288
 289 void *restart_stack __attribute__((__section__(".data")));
 290
 291 static void __init setup_lowcore(void)
 292 {
 293         struct _lowcore *lc;
 294
 295         /*
 296          * Setup lowcore for boot cpu
 297          */
 298         BUILD_BUG_ON(sizeof(struct _lowcore) != LC_PAGES * 4096);
 299         lc = __alloc_bootmem_low(LC_PAGES * PAGE_SIZE, LC_PAGES * PAGE_SIZE, 0);
 300         lc->restart_psw.mask = PSW_KERNEL_BITS;
 301         lc->restart_psw.addr =
 302                 PSW_ADDR_AMODE | (unsigned long) restart_int_handler;
 303         lc->external_new_psw.mask = PSW_KERNEL_BITS |
 304                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 305         lc->external_new_psw.addr =
 306                 PSW_ADDR_AMODE | (unsigned long) ext_int_handler;
 307         lc->svc_new_psw.mask = PSW_KERNEL_BITS |
 308                 PSW_MASK_DAT | PSW_MASK_IO | PSW_MASK_EXT | PSW_MASK_MCHECK;
 309         lc->svc_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) system_call;
 310         lc->program_new_psw.mask = PSW_KERNEL_BITS |
 311                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 312         lc->program_new_psw.addr =
 313                 PSW_ADDR_AMODE | (unsigned long) pgm_check_handler;
 314         lc->mcck_new_psw.mask = PSW_KERNEL_BITS;
 315         lc->mcck_new_psw.addr =
 316                 PSW_ADDR_AMODE | (unsigned long) mcck_int_handler;
 317         lc->io_new_psw.mask = PSW_KERNEL_BITS |
 318                 PSW_MASK_DAT | PSW_MASK_MCHECK;
 319         lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 320         lc->clock_comparator = -1ULL;
 321         lc->kernel_stack = ((unsigned long) &init_thread_union)
 322                 + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 323         lc->async_stack = (unsigned long)
 324                 __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
 325                 + ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 326         lc->panic_stack = (unsigned long)
 327                 __alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
 328                 + PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 329         lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 330         lc->thread_info = (unsigned long) &init_thread_union;
 331         lc->machine_flags = S390_lowcore.machine_flags;
 332         lc->stfl_fac_list = S390_lowcore.stfl_fac_list;
 333         memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list,
 334                MAX_FACILITY_BIT/8);
 335         if (MACHINE_HAS_VX)
 336                 lc->vector_save_area_addr =
 337                         (unsigned long) &lc->vector_save_area;
 338         lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 339         lc->sync_enter_timer = S390_lowcore.sync_enter_timer;
 340         lc->async_enter_timer = S390_lowcore.async_enter_timer;
 341         lc->exit_timer = S390_lowcore.exit_timer;
 342         lc->user_timer = S390_lowcore.user_timer;
 343         lc->system_timer = S390_lowcore.system_timer;
 344         lc->steal_timer = S390_lowcore.steal_timer;
 345         lc->last_update_timer = S390_lowcore.last_update_timer;
 346         lc->last_update_clock = S390_lowcore.last_update_clock;
 347
 348         restart_stack = __alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0);
 349         restart_stack += ASYNC_SIZE;
 350
 351         /*
 352          * Set up PSW restart to call ipl.c:do_restart(). Copy the relevant
 353          * restart data to the absolute zero lowcore. This is necessary if
 354          * PSW restart is done on an offline CPU that has lowcore zero.
 355          */
 356         lc->restart_stack = (unsigned long) restart_stack;
 357         lc->restart_fn = (unsigned long) do_restart;
 358         lc->restart_data = 0;
 359         lc->restart_source = -1UL;
 360
 361         /* Setup absolute zero lowcore */
 362         mem_assign_absolute(S390_lowcore.restart_stack, lc->restart_stack);
 363         mem_assign_absolute(S390_lowcore.restart_fn, lc->restart_fn);
 364         mem_assign_absolute(S390_lowcore.restart_data, lc->restart_data);
 365         mem_assign_absolute(S390_lowcore.restart_source, lc->restart_source);
 366         mem_assign_absolute(S390_lowcore.restart_psw, lc->restart_psw);
 367
 368 #ifdef CONFIG_SMP
 369         lc->spinlock_lockval = arch_spin_lockval(0);
 370 #endif
 371
 372         set_prefix((u32)(unsigned long) lc);
 373         lowcore_ptr[0] = lc;
 374 }
 375
 376 static struct resource code_resource = {
 377         .name  = "Kernel code",
 378         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 379 };
 380
 381 static struct resource data_resource = {
 382         .name = "Kernel data",
 383         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 384 };
 385
 386 static struct resource bss_resource = {
 387         .name = "Kernel bss",
 388         .flags = IORESOURCE_BUSY | IORESOURCE_MEM,
 389 };
 390
 391 static struct resource __initdata *standard_resources[] = {
 392         &code_resource,
 393         &data_resource,
 394         &bss_resource,
 395 };
 396
 397 static void __init setup_resources(void)
 398 {
 399         struct resource *res, *std_res, *sub_res;
 400         struct memblock_region *reg;
 401         int j;
 402
 403         code_resource.start = (unsigned long) &_text;
 404         code_resource.end = (unsigned long) &_etext - 1;
 405         data_resource.start = (unsigned long) &_etext;
 406         data_resource.end = (unsigned long) &_edata - 1;
 407         bss_resource.start = (unsigned long) &__bss_start;
 408         bss_resource.end = (unsigned long) &__bss_stop - 1;
 409
 410         for_each_memblock(memory, reg) {
 411                 res = alloc_bootmem_low(sizeof(*res));
 412                 res->flags = IORESOURCE_BUSY | IORESOURCE_MEM;
 413
 414                 res->name = "System RAM";
 415                 res->start = reg->base;
 416                 res->end = reg->base + reg->size - 1;
 417                 request_resource(&iomem_resource, res);
 418
 419                 for (j = 0; j < ARRAY_SIZE(standard_resources); j++) {
 420                         std_res = standard_resources[j];
 421                         if (std_res->start < res->start ||
 422                             std_res->start > res->end)
 423                                 continue;
 424                         if (std_res->end > res->end) {
 425                                 sub_res = alloc_bootmem_low(sizeof(*sub_res));
 426                                 *sub_res = *std_res;
 427                                 sub_res->end = res->end;
 428                                 std_res->start = res->end + 1;
 429                                 request_resource(res, sub_res);
 430                         } else {
 431                                 request_resource(res, std_res);
 432                         }
 433                 }
 434         }
 435 }
 436
 437 static void __init setup_memory_end(void)
 438 {
 439         unsigned long vmax, vmalloc_size, tmp;
 440
 441         /* Choose kernel address space layout: 2, 3, or 4 levels. */
 442         vmalloc_size = VMALLOC_END ?: (128UL << 30) - MODULES_LEN;
 443         tmp = (memory_end ?: max_physmem_end) / PAGE_SIZE;
 444         tmp = tmp * (sizeof(struct page) + PAGE_SIZE);
 445         if (tmp + vmalloc_size + MODULES_LEN <= (1UL << 42))
 446                 vmax = 1UL << 42;       /* 3-level kernel page table */
 447         else
 448                 vmax = 1UL << 53;       /* 4-level kernel page table */
 449         /* module area is at the end of the kernel address space. */
 450         MODULES_END = vmax;
 451         MODULES_VADDR = MODULES_END - MODULES_LEN;
 452         VMALLOC_END = MODULES_VADDR;
 453         VMALLOC_START = vmax - vmalloc_size;
 454
 455         /* Split remaining virtual space between 1:1 mapping & vmemmap array */
 456         tmp = VMALLOC_START / (PAGE_SIZE + sizeof(struct page));
 457         /* vmemmap contains a multiple of PAGES_PER_SECTION struct pages */
 458         tmp = SECTION_ALIGN_UP(tmp);
 459         tmp = VMALLOC_START - tmp * sizeof(struct page);
 460         tmp &= ~((vmax >> 11) - 1);     /* align to page table level */
 461         tmp = min(tmp, 1UL << MAX_PHYSMEM_BITS);
 462         vmemmap = (struct page *) tmp;
 463
 464         /* Take care that memory_end is set and <= vmemmap */
 465         memory_end = min(memory_end ?: max_physmem_end, tmp);
 466         max_pfn = max_low_pfn = PFN_DOWN(memory_end);
 467         memblock_remove(memory_end, ULONG_MAX);
 468
 469         pr_notice("Max memory size: %luMB\n", memory_end >> 20);
 470 }
 471
 472 static void __init setup_vmcoreinfo(void)
 473 {
 474         mem_assign_absolute(S390_lowcore.vmcore_info, paddr_vmcoreinfo_note());
 475 }
 476
 477 #ifdef CONFIG_CRASH_DUMP
 478
 479 /*
 480  * When kdump is enabled, we have to ensure that no memory from
 481  * the area [0 - crashkernel memory size] and
 482  * [crashk_res.start - crashk_res.end] is set offline.
 483  */
 484 static int kdump_mem_notifier(struct notifier_block *nb,
 485                               unsigned long action, void *data)
 486 {
 487         struct memory_notify *arg = data;
 488
 489         if (action != MEM_GOING_OFFLINE)
 490                 return NOTIFY_OK;
 491         if (arg->start_pfn < PFN_DOWN(resource_size(&crashk_res)))
 492                 return NOTIFY_BAD;
 493         if (arg->start_pfn > PFN_DOWN(crashk_res.end))
 494                 return NOTIFY_OK;
 495         if (arg->start_pfn + arg->nr_pages - 1 < PFN_DOWN(crashk_res.start))
 496                 return NOTIFY_OK;
 497         return NOTIFY_BAD;
 498 }
 499
 500 static struct notifier_block kdump_mem_nb = {
 501         .notifier_call = kdump_mem_notifier,
 502 };
 503
 504 #endif
 505
 506 /*
 507  * Make sure that the area behind memory_end is protected
 508  */
 509 static void reserve_memory_end(void)
 510 {
 511 #ifdef CONFIG_CRASH_DUMP
 512         if (ipl_info.type == IPL_TYPE_FCP_DUMP &&
 513             !OLDMEM_BASE && sclp.hsa_size) {
 514                 memory_end = sclp.hsa_size;
 515                 memory_end &= PAGE_MASK;
 516                 memory_end_set = 1;
 517         }
 518 #endif
 519         if (!memory_end_set)
 520                 return;
 521         memblock_reserve(memory_end, ULONG_MAX);
 522 }
 523
 524 /*
 525  * Make sure that oldmem, where the dump is stored, is protected
 526  */
 527 static void reserve_oldmem(void)
 528 {
 529 #ifdef CONFIG_CRASH_DUMP
 530         if (OLDMEM_BASE)
 531                 /* Forget all memory above the running kdump system */
 532                 memblock_reserve(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 533 #endif
 534 }
 535
 536 /*
 537  * Make sure that oldmem, where the dump is stored, is protected
 538  */
 539 static void remove_oldmem(void)
 540 {
 541 #ifdef CONFIG_CRASH_DUMP
 542         if (OLDMEM_BASE)
 543                 /* Forget all memory above the running kdump system */
 544                 memblock_remove(OLDMEM_SIZE, (phys_addr_t)ULONG_MAX);
 545 #endif
 546 }
 547
 548 /*
 549  * Reserve memory for kdump kernel to be loaded with kexec
 550  */
 551 static void __init reserve_crashkernel(void)
 552 {
 553 #ifdef CONFIG_CRASH_DUMP
 554         unsigned long long crash_base, crash_size;
 555         phys_addr_t low, high;
 556         int rc;
 557
 558         rc = parse_crashkernel(boot_command_line, memory_end, &crash_size,
 559                                &crash_base);
 560
 561         crash_base = ALIGN(crash_base, KEXEC_CRASH_MEM_ALIGN);
 562         crash_size = ALIGN(crash_size, KEXEC_CRASH_MEM_ALIGN);
 563         if (rc || crash_size == 0)
 564                 return;
 565
 566         if (memblock.memory.regions[0].size < crash_size) {
 567                 pr_info("crashkernel reservation failed: %s\n",
 568                         "first memory chunk must be at least crashkernel size");
 569                 return;
 570         }
 571
 572         low = crash_base ?: OLDMEM_BASE;
 573         high = low + crash_size;
 574         if (low >= OLDMEM_BASE && high <= OLDMEM_BASE + OLDMEM_SIZE) {
 575                 /* The crashkernel fits into OLDMEM, reuse OLDMEM */
 576                 crash_base = low;
 577         } else {
 578                 /* Find suitable area in free memory */
 579                 low = max_t(unsigned long, crash_size, sclp.hsa_size);
 580                 high = crash_base ? crash_base + crash_size : ULONG_MAX;
 581
 582                 if (crash_base && crash_base < low) {
 583                         pr_info("crashkernel reservation failed: %s\n",
 584                                 "crash_base too low");
 585                         return;
 586                 }
 587                 low = crash_base ?: low;
 588                 crash_base = memblock_find_in_range(low, high, crash_size,
 589                                                     KEXEC_CRASH_MEM_ALIGN);
 590         }
 591
 592         if (!crash_base) {
 593                 pr_info("crashkernel reservation failed: %s\n",
 594                         "no suitable area found");
 595                 return;
 596         }
 597
 598         if (register_memory_notifier(&kdump_mem_nb))
 599                 return;
 600
 601         if (!OLDMEM_BASE && MACHINE_IS_VM)
 602                 diag10_range(PFN_DOWN(crash_base), PFN_DOWN(crash_size));
 603         crashk_res.start = crash_base;
 604         crashk_res.end = crash_base + crash_size - 1;
 605         insert_resource(&iomem_resource, &crashk_res);
 606         memblock_remove(crash_base, crash_size);
 607         pr_info("Reserving %lluMB of memory at %lluMB "
 608                 "for crashkernel (System RAM: %luMB)\n",
 609                 crash_size >> 20, crash_base >> 20,
 610                 (unsigned long)memblock.memory.total_size >> 20);
 611         os_info_crashkernel_add(crash_base, crash_size);
 612 #endif
 613 }
 614
 615 /*
 616  * Reserve the initrd from being used by memblock
 617  */
 618 static void __init reserve_initrd(void)
 619 {
 620 #ifdef CONFIG_BLK_DEV_INITRD
 621         initrd_start = INITRD_START;
 622         initrd_end = initrd_start + INITRD_SIZE;
 623         memblock_reserve(INITRD_START, INITRD_SIZE);
 624 #endif
 625 }
 626
 627 /*
 628  * Check for initrd being in usable memory
 629  */
 630 static void __init check_initrd(void)
 631 {
 632 #ifdef CONFIG_BLK_DEV_INITRD
 633         if (INITRD_START && INITRD_SIZE &&
 634             !memblock_is_region_memory(INITRD_START, INITRD_SIZE)) {
 635                 pr_err("initrd does not fit memory.\n");
 636                 memblock_free(INITRD_START, INITRD_SIZE);
 637                 initrd_start = initrd_end = 0;
 638         }
 639 #endif
 640 }
 641
 642 /*
 643  * Reserve memory used for lowcore/command line/kernel image.
 644  */
 645 static void __init reserve_kernel(void)
 646 {
 647         unsigned long start_pfn = PFN_UP(__pa(&_end));
 648
 649 #ifdef CONFIG_DMA_API_DEBUG
 650         /*
 651          * DMA_API_DEBUG code stumbles over addresses from the
 652          * range [_ehead, _stext]. Mark the memory as reserved
 653          * so it is not used for CONFIG_DMA_API_DEBUG=y.
 654          */
 655         memblock_reserve(0, PFN_PHYS(start_pfn));
 656 #else
 657         memblock_reserve(0, (unsigned long)_ehead);
 658         memblock_reserve((unsigned long)_stext, PFN_PHYS(start_pfn)
 659                          - (unsigned long)_stext);
 660 #endif
 661 }
 662
 663 static void __init reserve_elfcorehdr(void)
 664 {
 665 #ifdef CONFIG_CRASH_DUMP
 666         if (is_kdump_kernel())
 667                 memblock_reserve(elfcorehdr_addr - OLDMEM_BASE,
 668                                  PAGE_ALIGN(elfcorehdr_size));
 669 #endif
 670 }
 671
 672 static void __init setup_memory(void)
 673 {
 674         struct memblock_region *reg;
 675
 676         /*
 677          * Init storage key for present memory
 678          */
 679         for_each_memblock(memory, reg) {
 680                 storage_key_init_range(reg->base, reg->base + reg->size);
 681         }
 682         psw_set_key(PAGE_DEFAULT_KEY);
 683
 684         /* Only cosmetics */
 685         memblock_enforce_memory_limit(memblock_end_of_DRAM());
 686 }
 687
 688 /*
 689  * Setup hardware capabilities.
 690  */
 691 static int __init setup_hwcaps(void)
 692 {
 693         static const int stfl_bits[6] = { 0, 2, 7, 17, 19, 21 };
 694         struct cpuid cpu_id;
 695         int i;
 696
 697         /*
 698          * The store facility list bits numbers as found in the principles
 699          * of operation are numbered with bit 1UL<<31 as number 0 to
 700          * bit 1UL<<0 as number 31.
 701          *   Bit 0: instructions named N3, "backported" to esa-mode
 702          *   Bit 2: z/Architecture mode is active
 703          *   Bit 7: the store-facility-list-extended facility is installed
 704          *   Bit 17: the message-security assist is installed
 705          *   Bit 19: the long-displacement facility is installed
 706          *   Bit 21: the extended-immediate facility is installed
 707          *   Bit 22: extended-translation facility 3 is installed
 708          *   Bit 30: extended-translation facility 3 enhancement facility
 709          * These get translated to:
 710          *   HWCAP_S390_ESAN3 bit 0, HWCAP_S390_ZARCH bit 1,
 711          *   HWCAP_S390_STFLE bit 2, HWCAP_S390_MSA bit 3,
 712          *   HWCAP_S390_LDISP bit 4, HWCAP_S390_EIMM bit 5 and
 713          *   HWCAP_S390_ETF3EH bit 8 (22 && 30).
 714          */
 715         for (i = 0; i < 6; i++)
 716                 if (test_facility(stfl_bits[i]))
 717                         elf_hwcap |= 1UL << i;
 718
 719         if (test_facility(22) && test_facility(30))
 720                 elf_hwcap |= HWCAP_S390_ETF3EH;
 721
 722         /*
 723          * Check for additional facilities with store-facility-list-extended.
 724          * stfle stores doublewords (8 byte) with bit 1ULL<<63 as bit 0
 725          * and 1ULL<<0 as bit 63. Bits 0-31 contain the same information
 726          * as stored by stfl, bits 32-xxx contain additional facilities.
 727          * How many facility words are stored depends on the number of
 728          * doublewords passed to the instruction. The additional facilities
 729          * are:
 730          *   Bit 42: decimal floating point facility is installed
 731          *   Bit 44: perform floating point operation facility is installed
 732          * translated to:
 733          *   HWCAP_S390_DFP bit 6 (42 && 44).
 734          */
 735         if ((elf_hwcap & (1UL << 2)) && test_facility(42) && test_facility(44))
 736                 elf_hwcap |= HWCAP_S390_DFP;
 737
 738         /*
 739          * Huge page support HWCAP_S390_HPAGE is bit 7.
 740          */
 741         if (MACHINE_HAS_HPAGE)
 742                 elf_hwcap |= HWCAP_S390_HPAGE;
 743
 744         /*
 745          * 64-bit register support for 31-bit processes
 746          * HWCAP_S390_HIGH_GPRS is bit 9.
 747          */
 748         elf_hwcap |= HWCAP_S390_HIGH_GPRS;
 749
 750         /*
 751          * Transactional execution support HWCAP_S390_TE is bit 10.
 752          */
 753         if (test_facility(50) && test_facility(73))
 754                 elf_hwcap |= HWCAP_S390_TE;
 755
 756         /*
 757          * Vector extension HWCAP_S390_VXRS is bit 11. The Vector extension
 758          * can be disabled with the "novx" parameter. Use MACHINE_HAS_VX
 759          * instead of facility bit 129.
 760          */
 761         if (MACHINE_HAS_VX)
 762                 elf_hwcap |= HWCAP_S390_VXRS;
 763         get_cpu_id(&cpu_id);
 764         add_device_randomness(&cpu_id, sizeof(cpu_id));
 765         switch (cpu_id.machine) {
 766         case 0x9672:
 767                 strcpy(elf_platform, "g5");
 768                 break;
 769         case 0x2064:
 770         case 0x2066:
 771         default:        /* Use "z900" as default for 64 bit kernels. */
 772                 strcpy(elf_platform, "z900");
 773                 break;
 774         case 0x2084:
 775         case 0x2086:
 776                 strcpy(elf_platform, "z990");
 777                 break;
 778         case 0x2094:
 779         case 0x2096:
 780                 strcpy(elf_platform, "z9-109");
 781                 break;
 782         case 0x2097:
 783         case 0x2098:
 784                 strcpy(elf_platform, "z10");
 785                 break;
 786         case 0x2817:
 787         case 0x2818:
 788                 strcpy(elf_platform, "z196");
 789                 break;
 790         case 0x2827:
 791         case 0x2828:
 792                 strcpy(elf_platform, "zEC12");
 793                 break;
 794         case 0x2964:
 795                 strcpy(elf_platform, "z13");
 796                 break;
 797         }
 798         return 0;
 799 }
 800 arch_initcall(setup_hwcaps);
 801
 802 /*
 803  * Add system information as device randomness
 804  */
 805 static void __init setup_randomness(void)
 806 {
 807         struct sysinfo_3_2_2 *vmms;
 808
 809         vmms = (struct sysinfo_3_2_2 *) alloc_page(GFP_KERNEL);
 810         if (vmms && stsi(vmms, 3, 2, 2) == 0 && vmms->count)
 811                 add_device_randomness(&vmms, vmms->count);
 812         free_page((unsigned long) vmms);
 813 }
 814
 815 /*
 816  * Setup function called from init/main.c just after the banner
 817  * was printed.
 818  */
 819
 820 void __init setup_arch(char **cmdline_p)
 821 {
 822         /*
 823          * print what head.S has found out about the machine
 824          */
 825         if (MACHINE_IS_VM)
 826                 pr_info("Linux is running as a z/VM "
 827                         "guest operating system in 64-bit mode\n");
 828         else if (MACHINE_IS_KVM)
 829                 pr_info("Linux is running under KVM in 64-bit mode\n");
 830         else if (MACHINE_IS_LPAR)
 831                 pr_info("Linux is running natively in 64-bit mode\n");
 832
 833         /* Have one command line that is parsed and saved in /proc/cmdline */
 834         /* boot_command_line has been already set up in early.c */
 835         *cmdline_p = boot_command_line;
 836
 837         ROOT_DEV = Root_RAM0;
 838
 839         /* Is init_mm really needed? */
 840         init_mm.start_code = PAGE_OFFSET;
 841         init_mm.end_code = (unsigned long) &_etext;
 842         init_mm.end_data = (unsigned long) &_edata;
 843         init_mm.brk = (unsigned long) &_end;
 844
 845         parse_early_param();
 846         os_info_init();
 847         setup_ipl();
 848
 849         /* Do some memory reservations *before* memory is added to memblock */
 850         reserve_memory_end();
 851         reserve_oldmem();
 852         reserve_kernel();
 853         reserve_initrd();
 854         reserve_elfcorehdr();
 855         memblock_allow_resize();
 856
 857         /* Get information about *all* installed memory */
 858         detect_memory_memblock();
 859
 860         remove_oldmem();
 861
 862         /*
 863          * Make sure all chunks are MAX_ORDER aligned so we don't need the
 864          * extra checks that HOLES_IN_ZONE would require.
 865          *
 866          * Is this still required?
 867          */
 868         memblock_trim_memory(1UL << (MAX_ORDER - 1 + PAGE_SHIFT));
 869
 870         setup_memory_end();
 871         setup_memory();
 872
 873         check_initrd();
 874         reserve_crashkernel();
 875         /*
 876          * Be aware that smp_save_dump_cpus() triggers a system reset.
 877          * Therefore CPU and device initialization should be done afterwards.
 878          */
 879         smp_save_dump_cpus();
 880
 881         setup_resources();
 882         setup_vmcoreinfo();
 883         setup_lowcore();
 884         smp_fill_possible_mask();
 885         cpu_init();
 886
 887         /*
 888          * Create kernel page tables and switch to virtual addressing.
 889          */
 890         paging_init();
 891
 892         /* Setup default console */
 893         conmode_default();
 894         set_preferred_console();
 895
 896         /* Setup zfcpdump support */
 897         setup_zfcpdump();
 898
 899         /* Add system specific data to the random pool */
 900         setup_randomness();
 901 }